Promotor, co-promotor, advisor : amin.shavandi@ulb.be, - , man.li@ulb.be

Research Unit : 3BIO-BIOMATTER

Description

attached pdf document

Promotor, co-promotor, advisor : amin.shavandi@ulb.be, - ,

Research Unit : 3BIO-BIOMATTER

Description

attached pdf document

Promotor, co-promotor, advisor : amin.shavandi@ulb.be, - , amir.abrishami@ulb.be

Research Unit : 3BIO-BIOMATTER

Description

attached pdf document

Promotor, co-promotor, advisor : amin.shavandi@ulb.be, - , Man.Li@ulb.be

Research Unit : 3BIO-BIOMATTER

Description

attached pdf document

Promotor, co-promotor, advisor : bernardo.innocenti@ulb.be, - , Bernardo Innocenti - Mattia Sisella

Research Unit : BEAMS - BIOMECHANICS

Description

Project title

Set-Up and Optimization of a Portable Knee Kinematic Simulator

Context

The Biomech research team at BEAMS has developed a portable knee kinematic simulator designed to replicate realistic joint movements in a variety of environments. In order to ensure its reliability and applicability in different clinical and experimental settings, it is essential to evaluate how variations in set-up parameters—such as camera positioning, calibration, and implant placement—affect the accuracy and repeatability of the measured kinematics.

Knee kinematics, particularly at the tibiofemoral and patellofemoral joints, are highly sensitive to changes in boundary conditions and measurement configurations. A systematic investigation is therefore necessary to identify the influence of these factors on kinematic output and to optimize the simulator for robust and reproducible use.

Objectives

To perform a sensitivity analysis of the effect of camera position, calibration parameters, and system settings on the accuracy and repeatability of knee kinematic measurements.

To assess how variations in implant positioning influence the reconstructed tibiofemoral and patellofemoral joint kinematics.

To propose an optimized set-up protocol that minimizes variability and improves reproducibility across different test environments.

Methods

Experimental Design: A series of trials will be conducted using the portable knee kinematic simulator under varied set-up conditions, including changes in camera angles, distances, calibration routines, and lighting environments.

Kinematic Analysis: Tibiofemoral and patellofemoral joint kinematics will be captured during physiologically realistic knee motion tasks using a motion tracking system. The data will be processed to extract key kinematic parameters such as flexion/extension, rotation, and translation.

Sensitivity Analysis: Statistical methods (e.g., ANOVA, regression models) will be used to quantify the effect of each variable on the kinematic output. Repeatability will be assessed through intra- and inter-session comparisons.

Optimization: Based on the findings, a set of recommendations and best practices for simulator configuration will be proposed to standardize future use in research or clinical validation.

Prerequisite

• Matlab,
• kinematic analisys

Contact person

For more information please contact : mattia.sisella@ulb.be

references

Bori et al, 2023, DOI: 10.1142/S0219519423400201 Arcidiaconi et al 2018, ISSN Print: 0976-6340 and ISSN Online: 0976-6359

Promotor, co-promotor, advisor : bernardo.innocenti@ulb.be, - , Bernardo Innocenti, Mattia Sisella

Research Unit : BEAMS - BIOMECHANICS

Description

Project title

Design of a Tibial Clamping Frame for In Vitro Investigation of Knee Kinematics Including Foot Interaction and Validation via Finite Element Modeling

Context

Accurate analysis of knee joint biomechanics is fundamental for understanding joint function in physiological and pathological conditions. Most in vitro studies isolate the knee and neglect the influence of distal segments, particularly the foot, which plays a significant role in load transmission and kinematic coupling during gait and functional movements.

To bridge this gap, there is a need for an experimental setup that maintains the anatomical and mechanical relationship between the tibia and foot while allowing precise control and measurement of joint motion. A custom-designed tibial clamping frame would enable stable fixation of cadaveric limbs, preserving the foot's contribution to motion and load distribution.

Moreover, coupling experimental measurements with finite element (FE) models allows for a deeper insight into internal stresses, contact mechanics, and joint interactions that are otherwise difficult to measure directly. Validation of FE models against realistic experimental data enhances their predictive value and clinical relevance. Many data are already available

Objective

To design and develop a tibial clamping frame for cadaveric testing that allows accurate investigation of knee kinematics, incorporating the mechanical effects of the foot.

To conduct experimental tests using the developed frame under passive and active loading conditions.

To create and validate a finite element model of the lower limb (tibia, knee, and foot complex) using experimental data.

Methods

Design and Prototyping: A tibial clamping frame will be designed to rigidly fix the tibia while allowing natural foot-ground interaction. The design will consider anatomical alignment, load distribution, and compatibility with existing motion capture and loading systems. The frame will be manufactured using metal and/or composite materials and tested for mechanical stability and usability.

Experimental Testing: Cadaveric lower limbs will be mounted in the frame, and tests will be conducted under both passive conditions (e.g., flexion-extension) and active loading (e.g., simulated muscle forces). Motion capture, force platforms, and pressure sensors will be used to record joint kinematics, ground reaction forces, and joint loading patterns.

Imaging and Geometry Acquisition: CT and/or MRI scans will be performed to obtain accurate geometries of bones, cartilage, and soft tissues, including the foot. These data will be used to reconstruct 3D models of the specimen.

Finite Element Modeling: A detailed FE model of the tibia-knee-foot complex will be developed based on the specimen-specific geometries. Material properties and boundary conditions will reflect those applied during experimental tests. The model will simulate joint mechanics and be validated against the recorded kinematics and force measurements.

Model Validation and Application: The FE model will be validated by comparing simulation results with experimental outputs (e.g., joint angles, contact pressures, load paths). The validated model will then be used to explore how changes in alignment, loading, or surgical interventions affect knee and foot mechanics.

Prerequisite

• Orthopaedic Biomechanics

Contact person

For more information please contact : mattia.sisella@ulb.be

references

Sisella et al, 2024 doi.org/10.1016/j.clinbiomech.2024.106353 Innocenti et al, 2020, DOI: 10.1016/j.knee.2019.09.007

Promotor, co-promotor, advisor : francois.quitin@ulb.be, - , Francois Quitin

Research Unit : BEAMS - EMBEDDED ELECTRONICS

Description

Evaluation of the energy consumption of the 5G Radio Access Network

Motivation

The increase of energy prices has prompted telecommunication operators to evaluate the impact of the energy consumption of their radio access network (RAN). For an operator, the RAN represents about 75% of their energy consumption, and is therefore a crucial element to reach a net zero carbon footprint. One of the main problems is that the consumption of the RAN is not proportional to the data traffic, as is illustrated in the figure below. Moreover, it is unclear which are the components of a base station that contribute strongly to the energy consumption in practice.

Objective This Master’s thesis will start from a database containing the measured energy consumption and data traffic of deployed 5G base stations in Belgium for two major operators. The student will first need to compare the energy consumption to the data traffic, and then parameterize the energy consumption based on the type of base station (number of sectors, number of antennas, number of frequency bands, …). The student will also construct a prediction model to project future energy consumption based on multiple deployment hypotheses.

Supervisors: Prof. François Quitin, Youssef Agram

Information : François Quitin (fquitin@ulb.be)

Students : ELEC, INFO, EM

Skills: data analysis, RAN deployment

attached pdf document

Promotor, co-promotor, advisor : francois.quitin@ulb.be, - , Francois Quitin

Research Unit : BEAMS - EMBEDDED ELECTRONICS

Description

Device localization with 5G 28 GHz mm-wave beamsteering arrays

Motivation

The 5G standard dedicated several gigahertz of bandwidth around 28 GHz for high-rate, low-range communications. To overcome the large path loss at millimeter-wave frequencies, large antenna arrays are required. We developed a pair of 28 GHz software-defined radios (SDRs), using conventional SDRs for generating the baseband signals, and a 16-antenna array for beamsteering. With our setup, signals with bandwidths up to 100 MHz can be generated, and the transmitter and receiver beams can be steered in azimuthal and elevation space through a digital control link. Since the transceiver’s beams are quite narrow, it is possible to identify the angle-of-arrival and angle-of-departure of the signal, which eventually allows to locate the transmitter node. High-accuracy localization of transmitters in 28 GHz is one of the many benefits of 28 GHz communications in 5G.

Objective This Master’s thesis will investigate the localization of a 28 GHz Tx in a beamsteering 28 GHz system. The main problem is that the process of changing the beam’s direction is not instantaneous, but rather takes between 1 and 10 milliseconds. Performing a full scan of all Tx and Rx directions is not a feasible option in practice. Therefore, different localization and tracking algorithms should be investigated (based on Kalman filters or particle filters) to allow for real-time tracking of moving transmitters. The algorithm could use only the array at the Rx side, or both Tx and Rx arrays. The proposed algorithms will be implemented and evaluated experimentally using the 28 GHz testbed, and the localization accuracy will be investigated.

Supervisors: Prof. François Quitin

Information : François Quitin (fquitin@ulb.be)

Students : ELEC, INFO, EM

Skills: Tracking filters, C++, experiment protocols

attached pdf document

Promotor, co-promotor, advisor : francois.quitin@ulb.be, - , Francois Quitin

Research Unit : BEAMS - EMBEDDED ELECTRONICS

Description

Design, implementation and testing of a CDMA-augmented Wireless CAN protocol

Motivation

The Controller Area Network (CAN) bus protocol is used for cabled broadcast networks in industrial and automation environments. It is favored because it guarantees low latency, prioritization of messages and heavy network loading. One of the unique features of CAN busses is it’s medium access control protocol, the CSMA/NDA protocol. Thanks to a concept of dominant and recessive bits on the communication bus, the CSMA/NDA implements a protocol that determines which nodes gets access to the medium without losing any time or bandwidth. It was commonly thought that CSMA/NDA could only be implemented in cabled networks, but we’ve recently proven that it is possible to define a wireless PHY layer protocol that is compatible with the CSMA/NDA protocol of the MAC layer. In a proof-of-concept experiment, we designed and tested wireless transceivers that are fully compatible with the CAN controller of our microcontroller systems, proving the feasibility of wireless CAN (WiCAN) communications.

Objective This Master’s thesis will start from our proof-of-concept experiment, and provide a more robust version of our WiCAN transceivers by introducing a layer of CDMA (code division multiple access) over the CAN protocol, allowing for more robust communications.

Supervisors: Prof. François Quitin, Michel Osée

Information : François Quitin (fquitin@ulb.be)

Students : ELEC, INFO, EM

Skills: Electronics, PCB design, industrial networks

attached pdf document

Promotor, co-promotor, advisor : francois.quitin@ulb.be, - , Francois Quitin

Research Unit : BEAMS - EMBEDDED ELECTRONICS

Description

Environment Mapping with 5G 28 GHz mm-wave beamsteering arrays

Motivation

The 5G standard dedicated several gigahertz of bandwidth around 28 GHz for high-rate, low-range communications. To overcome the large path loss at millimeter-wave frequencies, large antenna arrays are required. We developed a pair of 28 GHz software-defined radios (SDRs), using conventional SDRs for generating the baseband signals, and a 16-antenna array for beamsteering. With our setup, signals with bandwidths up to 100 MHz can be generated, and the transmitter and receiver beams can be steered in azimuthal and elevation space through a digital control link. Since the transceiver’s beams are quite narrow, it is possible to identify individual multipath components in the environment, and eventually to recreate a map of the obstacles in the environment by steering the Tx and Rx beam in all directions to “scan” the environment.

Objective

This Master’s thesis will investigate the ability of a 28 GHz beamsteering system to resolve multipath components and eventually reconstruct an image of the environment surrounding our transceivers. The first step will be to adapt the SDR software to allow for transmitting and receiving signals up to 100 MHz of bandwidth. Next, the student will need to design and implement super-resolution algorithms, such as the expectation-maximization algorithm, to resolve individual multipath contributions in the received signal. Such high-complexity mathematical algorithms have been successfully leveraged for multipath identification in channel sounding scenarios, but must be adapted to account for the process of beamsteering in the proposed system. The student will have to perform measurement in synthetic and non-synthetic scenarios, and validate the super-resolution algorithm on these measurements.

Supervisors: Prof. François Quitin, Nigus Yirga

Information : François Quitin (fquitin@ulb.be)

Students : ELEC, PHYS, INFO

Skills: super-resolution algorithms, C++, experiment protocols

attached pdf document

Promotor, co-promotor, advisor : bernardo.innocenti@ulb.be, - , Bernardo Innocenti, Mattia Sisella

Research Unit : BEAMS BIOMECHANICS

Description

Project title

Development and Validation of a Finite Element Model of the Patellofemoral Joint

Context

The patellofemoral joint plays a crucial role in knee function, particularly in load distribution, force transmission, and joint stability during both passive and active motion. Due to its complex geometry and contact mechanics, computational models—especially finite element (FE) models—are increasingly used to investigate patellofemoral biomechanics, guide surgical planning, and design orthopaedic implants.

Despite advances in computational modeling, validated FE models that accurately replicate the dynamic behavior of the patellofemoral joint remain scarce. To improve model fidelity and clinical relevance, validation against experimental data from cadaveric testing under physiological conditions is essential.

This thesis is in collaboration with Dr. Gianluca Castellarin

Objective

To develop a high-fidelity finite element model of the human patellofemoral joint based on medical imaging and anatomical data.

To use experimentally acquired patellofemoral kinematics and contact mechanics of cadaveric specimens during both passive and active knee motions.

To validate the FE model against experimental measurements and assess its predictive capabilities under various loading conditions.

Methods

Imaging and Geometry Reconstruction: High-resolution CT and MRI scans of each specimen will be used to segment and reconstruct 3D geometries of the femur, patella, cartilage, and relevant soft tissues (e.g., quadriceps tendon, patellar tendon, retinacula).

Finite Element Modeling: An FE model of the patellofemoral joint will be built using the reconstructed geometries. Material properties will be assigned based on literature data or mechanical testing. Contact definitions and boundary conditions will replicate those used in the cadaveric experiments.

Model Validation: Simulated joint kinematics, contact forces, and pressure distributions will be compared with experimental data. Sensitivity analyses will be performed to evaluate the influence of model assumptions (e.g., material properties, boundary conditions) on output accuracy.

Applications: Once validated, the model may be used to simulate pathological conditions (e.g., maltracking, patellar instability) or evaluate the biomechanical performance of implants and surgical interventions.

Prerequisite

• Orthopaedic Biomechanics,

Contact person

For more information please contact : mattia.sisella@ulb.be

references

Innocenti et al, 2020 DOI: 10.1016/j.knee.2019.09.007 Sisella, et al, 2024, doi.org: 10.1016/j.clinbiomech.2024.106353

Promotor, co-promotor, advisor : bernardo.innocenti@ulb.be, - , Bernardo Innocenti

Research Unit : BEAMS BIOMECHANICS

Description

Project title

Biomechanical Analysis of Bone-Implant Interaction in Pediatric Proximal Humerus Reconstructions with Tumoral Shoulder Implants: Effect of Resection Level and Fixation Technique

Context

Limb-sparing surgery is increasingly used in the treatment of pediatric bone tumors, particularly in the proximal humerus, where osteosarcoma and Ewing sarcoma are most frequently located. The use of tumoral shoulder implants enables preservation of limb function and quality of life, but pediatric applications present additional challenges due to ongoing skeletal growth, reduced bone size and density, and longer life expectancy, which heighten the risk of implant failure or revision.

Optimizing implant fixation strategies for pediatric patients requires a detailed understanding of the biomechanical interaction between the implant and the remaining bone. Critical factors include the level of bone resection, the choice between cemented and press-fit fixation, and the type of mechanical stabilization used (e.g., intramedullary stem, lateral plate, or both). Finite element (FE) modeling is an essential tool to evaluate these parameters in silico, offering predictive insight into bone-implant mechanics tailored to pediatric anatomy and clinical constraints.

Objective

To develop anatomically accurate finite element models of pediatric proximal humerus reconstructions using tumoral shoulder implants.

To evaluate how resection level and fixation technique (cemented vs. press-fit; stem only vs. lateral plate vs. combined) affect:

• Bone stress distribution,
• Implant stability,
• Risk of loosening or failure.

To provide evidence-based guidelines for surgical planning and implant design in pediatric oncologic shoulder reconstructions.

Methods

Anatomical Modeling: A pediatric humerus model will be developed from anonymized CT or MRI data of children in the appropriate age range. Several resection scenarios (short, medium, long) will be simulated to reflect typical tumor excisions.

Implant Configuration: The reconstructed models will include: - Tumoral shoulder implants scaled to pediatric dimensions, - Fixation variants: press-fit vs. cemented stem; lateral plate; and combination fixation. - Material Properties: Pediatric-specific bone material properties (cortical and trabecular) will be defined based on literature or published databases. Implant and cement properties will be modeled as nonlinear or elastic-plastic where appropriate. - Loading and Boundary Conditions: Simulated loading scenarios will include shoulder abduction, flexion, and functional loads typical for pediatric rehabilitation. Growth-related bone remodeling will not be directly modeled, but its implications will be discussed.

Analysis Parameters: - Stress and strain distribution in residual bone, - Bone-implant micromotion, - Predicted zones of high failure risk or stress shielding, - Influence of fixation strategy on mechanical stability.

Sensitivity Analyses: Investigate how variations in bone density, implant fit, or surgical alignment affect mechanical outcomes in the pediatric context.

Prerequisite

• Orthopaedic Biomechanics

Contact person

For more information please contact : bernardo.innocenti@ulb.be

references

upon request to Prof. Innocenti bernardo.innocenti@ulb.be

Promotor, co-promotor, advisor : alain.delchambre@ulb.be, Pierre Lambert, Pascal Doguet and Simon Den Haene

Research Unit : BEAMS BIOMECHATRONICS

Description

Project title

The project aims to design and develop a nebulization catheter to address respiratory distress in premature babies

Context

Preterm infants, especially those born before 34 weeks of gestation, are at high risk of developing Respiratory Distress Syndrome (RDS). RDS occurs due to insufficient surfactant, leading to difficulty in breathing and inadequate oxygenation of the blood. Surfactant is a substance that is critical for the proper function of the lungs, particularly in newborns. It is a mixture of lipids and proteins produced naturally in the lungs and works to lower surface tension within the alveoli, the tiny air sacs where gas exchange occurs. By reducing surface tension, surfactant prevents the alveoli from collapsing and allows them to remain open during exhalation, facilitating efficient gas exchange and making breathing easier. Without enough surfactant, the alveoli tend to collapse after each breath, making it extremely hard for the infant to keep their lungs inflated. Administering surfactant helps keep the alveoli open, allowing for more efficient and less laborious breathing. The administration can be done soon after birth (prophylactic use) or when the symptoms of RDS become apparent (rescue therapy). The classical technique to administer surfactant involves the insertion of an endotracheal tube by means of laryngoscope, followed by the insertion of a smaller catheter with the surfactant inside the tube. This technique called INSURE, which stands for INtubation, SURfactant administration, and Extubation, is invasive, requires sedation due to the discomfort and also brings risks of physical trauma to the airway such as laryngeal and tracheal injury, vocal cord damage, bleeding or rapid extubation failure [2]. Other techniques were developed to overcome these drawbacks [3], [4], [5], [6]: the LISA technique still uses a laryngoscope followed by administration with a dedicated catheter but can still be a source of pain, especially when performed without any anesthesia. A newly developed technique, called FAST [7] uses a less invasive approach by inserting a slim endoscope, as shown on Fig. 3, through the nose and delivering the surfactant at the laryngeal level by means of a sub millimeter catheter inserted through the working channel of the endoscope. The FAST technique also brings the advantage of being able to keep the baby on CPAP (continuous positive airway pressure) while administering the surfactant. Another approach would be to use nebulization. Surfactant delivery by nebulization would be even less invasive. Some tries have already been done [2][8], delivering nebulized surfactant at the CPAP mask level but surfactant delivery in the lungs is not effective, with a few percents of the volume reaching the target. Therefore, there is a need to go further and enhance the technique.

Objective

This master thesis aims to study and design a nebulization catheter to be inserted in the pharyngeal area without the need of endoscopy nor sedation. Nebulization will be performed at the tip of the catheter by mixing compressed air and surfactant. By means of a pressure sensor, nebulization can also be synchronized on the breathing pattern for effective delivery to the lungs. The main goal of this thesis is to study, analyze and design the nebulization part of the device. The nebulization occurs at the tip of the catheter where air and liquid are mixed together to create drops of appropriate size. This thesis is a continuation of a thesis during which the mixing pneumatic set-up has been developed and tested with an initial 3D printed catheter tip based on examples of the literature. It now needs to be further brought to the final product. Requirements in terms of pressure, volume, droplets size would need to be well defined. Different types of tips will be investigated and fabricated as control of mixing variables will be studied. This work will also be done in close collaboration with medical doctors. The catheter tip will use the advantages brought by 3D printing technologies to further enhance the state of the art and improve mixing and delivery. This work can also involve the management of delivery and pressure synchronization to the breathing pattern, which is the goal of another MFE. In this case, it would require some collaboration with the other student.

Contact person

Pascal Doguet (pascal.doguet@ulb.be) and Simon Den Haene (simon.den.haene@ulb.be)

references

[1] Onybiotech, “Onybiotech - what is RDS ?” [2] D. G. Sweet et al., “European Consensus Guidelines on the Management of Respiratory Distress Syndrome - 2019 Update,” Neonatology, vol. 115, no. 4. S. Karger AG, pp. 432–450, Jun. 01, 2019. doi: 10.1159/000499361. [3] M. K. Sabzehei, B. Basiri, M. Shokouhi, S. Ghahremani, and A. Moradi, “Comparison of minimally invasive surfactant therapy with intubation surfactant administration and extubation for treating preterm infants with respiratory distress syndrome: a randomized clinical trial,” Clin Exp Pediatr, vol. 65, no. 4, pp. 188–193, Apr. 2022, doi: 10.3345/cep.2021.00297. [4] G. H. Shim, “Update of minimally invasive surfactant therapy,” Korean J Pediatr, vol. 60, no. 9, pp. 273–281, Sep. 2017, doi: 10.3345/kjp.2017.60.9.273. [5] “Verder et alSurfactant Therapy & Nasal CPAP for Newborns with Respiratory Distress Syndrome1994”. [6] E. Herting, C. Härtel, and W. Göpel, “Less invasive surfactant administration: best practices and unanswered questions,” Current Opinion in Pediatrics, vol. 32, no. 2. Lippincott Williams and Wilkins, pp. 228–234, Apr. 01, 2020. doi: 10.1097/MOP.0000000000000878. [7] D. Guevorkian et al., “Pilot Study to Evaluate a New Method for Endotracheal Administration of Surfactant in Neonatal Respiratory Distress Syndrome: Fiberscope Assisted Surfactant Therapy (FAST),” Neonatology, vol. 119, no. 6, pp. 753–759, Dec. 2022, doi: 10.1159/000525848. [8] J. J. Pillow and S. Minocchieri, “Innovation in surfactant therapy II: Surfactant administration by aerosolization,” Neonatology, vol. 101, no. 4. pp. 337–344, Jun. 2012. doi: 10.1159/000337354. [9] A. Tronde, G. Baran, S. Eirefelt, H. Lennernäs, and U. H. Bengtsson, “Miniaturized Nebulization Catheters: A New Approach for Delivery of Defined Aerosol Doses to the Rat Lung,” 2002. [10] “US20090107503A1_Nebulizing catheter system and methods of use and manufacture”

Promotor, co-promotor, advisor : jean-francois.determe@ulb.be, - , Jean-François Determe

Research Unit : BEAMS-EE

Description

Project title

Hybrid CPU-GPU computing for sparse signal processing

Context

Sparse signal processing is about exploiting the sparse nature of signals to process them more efficiently or to measure them to some prescribed level of accuracy with fewer measurements than those conventional methods require [1]. The idea is that we would like to retrieve some signal belonging to a high dimensional space while leveraging its sparse nature so as to lower the number of measurements needed to reconstruct it in comparison to traditional approaches (which require at least as many measurements as space dimensions). Among the classes of sparse recovery algorithms, greedy ones are particularly important for embedded applications because of their low space and time complexity (in comparison to other alternatives based on convex optimization problems) [2]. A particular class of greedy algorithms is that tailored to multiple measurement vector (MMV) problems: several signals are acquired by different acquisition channels, which may be different but, by their nature, share the same support (i.e., the position of their non-zero entries are identical or, at least, similar). These problems appear in many practical applications, including some where algorithms are implemented on embedded platforms with limited computational capabilities [3-6].

There has been some research on the computationally optimal implementation of greedy algorithms such as orthogonal matching pursuit (OMP). Some of this research is mathematical (see, e.g., [2]) and aims to find algebraic ways of recasting the original OMP algorithm into one with lower complexity in time and space. Other avenues of optimization focus on how to optimally implement such algorithms on hardware platforms: regular processors, processors augmented with graphical processing units (GPUs), field programmable gate arrays (FPGAs) (e.g., [7]) and application-specific integrated circuits (ASICs) [8]. Surprisingly, despite their high costs, most research on the subject has focused on FPGA and ASIC-based architectures (see, e.g., [9, Sec. 4.2, Table 3] for a recent review).

Objective and steps

Loosely speaking, this master’s thesis studies how to best implement OMP, particularly in multiple measurement vector (MMV) scenarios. In particular, the master’s thesis student is expected to evaluate to what extent GPU-based computations are faster or slower than those on ARM or x86_64 processors, depending on the number of measurement vectors in the MMV problem. Steps are: i) to review mathematically optimized implementations of OMP, ii) to program a naïve OMP in Python or Matlab to get an idea of how it works, iii) to implement OMP for MMV problems using BLAS/LAPACK (with an MKL or OpenBLAS implementation) and/or Eigen, iv) to implement OMP for MMV problems using CUDA (and libraries such as cuBLAS and cuSOLVER when appropriate) and v) to compare results against the naïve Python/Matlab implementation and other existing implementations if they are available as well as to determine the extent to which GPU-based algorithms run faster or slower than those limited to running on CPUs only. The BEAMS-EE department possesses various CPU-GPU platforms that the student can use (ranging from embedded devices to a server rack with dual GPUs).

Student profile

This subject is appropriate for students willing to translate mathematical algorithms into efficient implementations running on modern computing architectures. It requires some skills in linear algebra, computer architecture and C/C++/CUDA programming. Having followed the course “Microprocessor architecture” and/or a course on GPU programming is a plus but is not mandatory. Having experience with BLAS/LAPACK is a plus as well.

Referrences

[1] Donoho, D. L. (2006). Compressed sensing. IEEE Transactions on information theory, 52(4), 1289-1306. [2] Sturm, B. L., & Christensen, M. G. (2012, August). Comparison of orthogonal matching pursuit implementations. In 2012 Proceedings of the 20th European Signal Processing Conference (EUSIPCO) (pp. 220-224). IEEE. [3] Mishali, M., Eldar, Y. C., & Elron, A. J. (2011). Xampling: Signal acquisition and processing in union of subspaces. IEEE Transactions on Signal Processing, 59(10), 4719-4734. [4] Mishali, M., & Eldar, Y. C. (2009). Blind multiband signal reconstruction: Compressed sensing for analog signals. IEEE Transactions on signal processing, 57(3), 993-1009. [5] Shahmansoori, A., Garcia, G. E., Destino, G., Seco-Granados, G., & Wymeersch, H. (2017). Position and orientation estimation through millimeter-wave MIMO in 5G systems. IEEE Transactions on Wireless Communications, 17(3), 1822-1835. [6] Zhou, Y., Song, A., Tong, F., & Kastner, R. (2018). Distributed compressed sensing based channel estimation for underwater acoustic multiband transmissions. The Journal of the Acoustical Society of America, 143(6), 3985-3996. [7] Huang, G., & Wang, L. (2017). An FPGA-based architecture for high-speed compressed signal reconstruction. ACM Transactions on Embedded Computing Systems (TECS), 16(3), 1- 23. [8] Chen, K. T., Ma, W. H., Hwang, Y. T., & Chang, K. Y. (2020). A low complexity, high throughput DoA estimation chip design for adaptive beamforming. Electronics, 9(4), 641. [9] Djelouat, H., Amira, A., & Bensaali, F. (2018). Compressive sensing-based IoT applications: A review. Journal of Sensor and Actuator Networks, 7(4), 45.

Contact

Jean-François Determe, jean-francois.determe@ulb.be Solbosch campus, building U, level. 5, BEAMS-EE department

attached pdf document

Promotor, co-promotor, advisor : jean-francois.determe@ulb.be, - , Jean-François Determe

Research Unit : BEAMS-EE

Description

Project title

Accelerating autoregressive integrated moving average models using sparse matrix representations

Context

Statistical time series models (such as autoregressive integrated moving average, ARIMA) are a staple forecasting method for time series. Even with the advent of more modern forecasting methods (e.g., long-short term memory (LSTM) neural networks and Facebook Prophet), they remain an interesting baseline that has the advantage of being interpretable.

Modern ARIMA implementations rely on Kalman filters for fitting and forecasting purposes (notably because Kalman filters and state-space representations are well mastered mathematical tools and support missing observations). This is also the case for seasonal time series models (i.e., seasonal autoregressive integrated moving average, SARIMA) [Brockwell, 2002] [Durbin, 2012].

Interestingly, state-space models of high-order ARIMA models happen to be sparse (which means that the state transition and observation matrices mostly consist of zero coefficients); sparsity levels get even higher in the case of SARIMA models.

However, despite the sparse nature of most of the state-space matrices, it appears that many implementations (e.g., Python statsmodels) do not leverage it and instead rely on standard dense matrix representations for computations.

Objective and steps

The goal of this master’s thesis is to program a custom, high-performance fit-and-forecast program for ARIMA models, relying on sparse matrix representations to reduce computing load. Steps are i) a review of state-space models for ARIMA models and the associated fitting and forecasting methods, ii) implementing a fit-and-forecast program that leverages sparse matrix representations and iii) compare the performance of the developed program to that of Python’s statsmodels (in terms of accuracy and computing time), using real and/or synthetic time series. Very motivated students can go further and i) extend the program to SARIMA models, or ii) implement the program in CUDA (for graphical processing unit (GPU) programming) or iii) rely on template meta-programming for implementing compile-time sparse matrices. The BEAMS-EE department possesses various computing platforms (notably a consumer-grade computing tower with a GPU and a rack server with dual Xeon Gold (64 cores in total) and two RTX A6000 GPUs), which are available to the student.

Student profile

Ideally, the student has experience in C++ programming and is skilled in mathematics (mostly linear algebra, although basic probability theory and optimization theory are important as well). Knowledge of BLAS, LAPACK and/or the Eigen library is a plus. Having followed the course “Microprocessor architectures” is also a (minor) plus.

Referrences

[Brockwell, 2002] Brockwell, Peter J., and Richard A. Davis, eds. Introduction to time series and forecasting. New York, NY: Springer New York, 2002. [Durbin, 2012] Durbin, James, and Siem Jan Koopman. Time series analysis by state space methods. Vol. 38. OUP Oxford, 2012.

Contact

Jean-François Determe, jean-francois.determe@ulb.be Solbosch campus, building U, level. 5, BEAMS-EE department

attached pdf document

Promotor, co-promotor, advisor : jean-francois.determe@ulb.be, - , Jean-François Determe

Research Unit : BEAMS-EE

Description

Project title

Adaptive modulations for underwater acoustic communications

Context

The use of Adaptive Modulations (AM) for underwater acoustic communication is a recent research endeavour, whose principle is to dynamically adapt modulation parameters (e.g., the modulation order M of a M-PSK or M-FSK modulations) or even switch between fundamentally different modulations (e.g., switch from QPSK to 2-FSK). The switch from one modulation to another depends on environmental data (e.g., bathymetry data) and communication data (e.g., estimated signal-to-noise ratio and channel impulse response) and is triggered in such a way that the “best” modulation is chosen at any given time; if correctly designed, adaptive modulation should translate into throughputs that are higher than those obtained with a fixed modulation.

A few representative papers about adaptive modulations are [Pelekanis, 2018] (based on boosted regression trees), [Huang, 2020] (based on support vector machines), [Kojima, 2021] (based on convolutional neural networks) and [Su, 2019] as well as [Cui, 2023)] (based on reinforcement learning).

Given that underwater acoustic channels are complex and that real-world experiments are difficult to carry out, researchers increasingly rely on modern acoustic propagation simulators (e.g., the world oceanic simulation system (WOSS) [Casari, 2014]).

Objective and steps

The main goal of the master’s thesis is to implement (and possibly extend), simulate and compare two representative adaptive modulation approaches. The steps are i) to setup WOSS in a Debian/Ubuntu-based operating system (computing servers are available for that purpose) and get familiar with it, ii) to implement (at least) two modern adaptive modulation algorithms on WOSS (extending them to a wider breadth of modulations than originally designed for is recommended for motivated students) and iii) to investigate which adaptive modulation methods perform best and in which circumstances (shallow vs deep waters, short-range vs long-range communication, etc.)

Student profile

Although not mandatory, the student should ideally have followed the courses “Modulation and coding” and “Communication channels”. Having followed or following a course on machine learning (e.g., “Machine Learning and Big Data Processing”) is also a plus. The subject is particularly suited to students willing to apply modern machine learning to telecommunications problems.

References

[Casari, 2014] Casari, Paolo et al. "Open source suites for underwater networking: WOSS and DESERT underwater." IEEE Network 28, no. 5 (2014): 38-46. [Cui, 2023] Cui, Xuerong et al. "Reinforcement learning-based adaptive modulation scheme over underwater acoustic OFDM communication channels." Physical Communication 61 (2023): 102207. [Huang, 2020] Huang, Jianchun, and Roee Diamant. "Adaptive modulation for long-range underwater acoustic communication." IEEE Transactions on Wireless Communications 19, no. 10 (2020): 6844-6857. [Kojima, 2021] Kojima, Shun et al. "CNN-based joint SNR and Doppler shift classification using spectrogram images for adaptive modulation and coding." IEEE Transactions on Communications 69, no. 8 (2021): 5152-5167. [Pelekanis, 2018] Pelekanakis, Konstantinos, and Luca Cazzanti. "On adaptive modulation for low SNR underwater acoustic communications." In OCEANS 2018 MTS/IEEE Charleston, pp. 1-6. IEEE, 2018. [Su, 2019] Su, Wei et al. "Reinforcement learning-based adaptive modulation and coding for efficient underwater communications." IEEE access 7 (2019): 67539-67550.

Contact

Jean-François Determe, jean-francois.determe@ulb.be Solbosch campus, building U, level. 5, BEAMS-EE department

attached pdf document

Promotor, co-promotor, advisor : jean-francois.determe@ulb.be, - , Jean-François Determe

Research Unit : BEAMS-EE

Description

Project title

Design of an underwater acoustic modem

Context

Underwater communications are peculiar in that traditional electromagnetic communications are inappropriate beyond a few meters due to high attenuation in water. As a result, all medium and long-range underwater communications are acoustic (i.e., they use pressure waves). Bandwidths are also typically orders of magnitude lower than those of over-the-air wireless communications (in underwater acoustic communications, bandwidth generally does not exceed 20 kHz and the carrier frequency is often between 10 and 40 kHz) [Zia, 2021]. These low bandwidths led to the development of many research and commercial modems based on micro-controller units (MCUs) or microprocessor units (MPUs), with some high-end modems relying on field-programmable gate arrays (FPGAs) instead. Given the low carrier frequency, no analog upconversion and downconversion are needed and signals are generally directly sampled (RX) or converted to analog (TX) in bandpass.

Objective and steps

The goal of this master’s thesis is to develop the digital parts of an underwater acoustic modem. Several analog parts (power amplifier at the TX and low-noise conditioning chain at the RX) shall not be developed given how difficult they are to validate. Steps are i) to develop a MCU-based TX supporting a simple modulation (e.g., BPSK) without error correction features ii) to develop a MCU-based RX that demodulates the chosen simple modulation (including basic channel equalization) and to validate it using the TX (a direct wired connection connects the TX to the RX) iii) (optional) to implement the TX and/or RX on a custom printed circuit board (PCB) iv) (optional) to implement error correction with convolutional codes on the TX and RX. A high-end STM32H7 MCU is expected to be used for implementing both the TX and the RX. Evaluation/Development boards for the STM32H7 MCU will be available to the student. The student is expected to leverage microarchitecture enhancements to optimize performance. The student may rely on existing open-source modems (e.g., that of [Renner, 2020], which is based on the similar but less powerful STM32F4 MCU).

Student profile

Ideally, the student should have followed the courses “Microprocessor architecture” and “Modulation and coding”. This master’s thesis is geared toward students that have an interest in building real telecommunication devices and are eager to learn how to carry out practical work on modern, advanced MCU architectures.

References

[Renner, 2020] Renner, Bernd-Christian et al. "AHOI: Inexpensive, low-power communication and localization for underwater sensor networks and μAUVs." ACM Transactions on Sensor Networks (TOSN) 16, no. 2 (2020): 1-46. [Zia, 2021] Zia, Muhammad Yousuf Irfan et al. "State-of-the-art underwater acoustic communication modems: Classifications, analyses and design challenges." Wireless personal communications 116 (2021): 1325-1360.

Contact

Jean-François Determe, jean-francois.determe@ulb.be Solbosch campus, building U, level. 5, BEAMS-EE department

attached pdf document

Promotor, co-promotor, advisor : alain.delchambre@ulb.be, - , Charlotte Deroubaix

Research Unit : BEAMS/BIOMECHATRONICS

Description

Project title

The project aims to develop a finite element model of a layer jamming-based variable sttiffness mechanism for flexible endoscopic structures

Context

Minimally invasive surgery using endoscopic techniques is known to reduce complications during operations, promoting faster patient recovery, lower healthcare costs, and improved access to care. Technological advancements have further driven the growth of interventional endoscopy, with an annual average of 432,271 upper digestive endoscopic procedures performed in Belgium between 2013 and 2023 [1]. However, endoscopes are a known source of nosocomial infections due to their complex structure and internal channels. Traditional endoscopes consist of a long insertion tube with a lighting and visualization system, a working channel for surgical tools, and additional channels for rinsing, aspiration, or insufflation [2] . Their distal tip is controlled by cable-actuated mechanisms. Despite strict hospital cleaning protocols, some areas remain difficult to sterilize, allowing pathogens to persist and cause cross-contamination [3]. In 2019, these safety concerns led the U.S. Food and Drug Administration (FDA) to recommend using partially or fully disposable devices, initiating a shift toward single-use endoscopes [4]. While disposable endoscopes reduce infection risks, they pose significant environmental, economic, and social challenges. A recent study [5] estimated that U.S. hospitals produce around 38,000 tons of waste annually from endoscopic procedures, a number that would increase by 40% if all procedures relied solely on single-use endoscopes. Of the 35% of recyclable endoscopic waste, only 9% is actually recycled, while the rest is incinerated (12%) or sent to landfills (79%). Another study [6] found that these devices generate approximately 15.78 tons of CO₂ emissions annually in the U.S. Beyond environmental concerns, disposable endoscopes are also costly in the long run [7]. Although their initial purchase price is lower, high usage volumes drive up expenses. To cut costs, manufacturers often reduce functionalities, potentially compromising performance compared to reusable models. Given these limitations, single-use devices in their current form are unsustainable. To overcome these challenges, the BEAMS is developing a semi-disposable endoscope based on eco-friendly pneumatic actuation.This design will reduce waste by making only the insertion tube disposable (not possible with cable-actuated systems) while the control handle and workstation will remain reusable after basic cleaning, as they do not contact the patient. Additionally, the tube's design will optimize material use, enhance recyclability, and incorporate modularity for easier maintenance and repairs [8]. Beyond environmental benefits, pneumatic actuation can improve navigation through complex anatomical pathways while lowering the risk of patient injury, and enhance endoscopic localization capabilities by using MRI-compatible materials [9]. A previous study conducted within the BEAMS research group demonstrated the feasibility of this approach with a functional prototype [10]: a 90 cm-long, 5 mm-diameter steerable catheter made from a single material and capable of bending in all directions.

Objective

The soft nature of materials used for pneumatic actuation of flexible proximal sections in endoscopic devices creates certain limitations. Specifically, procedures such as biopsies are difficult to perform because the overly compliant structure does not allow for sufficient force application and lacks the stability necessary for precise operations. These limitations compromise the efficiency and accuracy of endoscopic interventions. One promising solution to this limitation is integrating a variable stiffness mechanism into the flexible proximal section of the endoscopic device. Layer jamming is a particularly suitable approach. This technique modulates stiffness by compressing a stack of thin, flexible layers enclosed within an airtight membrane. When vacuum pressure is applied, atmospheric pressure compresses the layers together, increasing friction and significantly enhancing the structure's overall rigidity. Controlling the level of vacuum allows the stiffness to be adjusted in real time, enabling the structure to transition between a soft, compliant state and a stiffer one, more stable configuration based on procedural needs. The objectives of this master thesis would be: 1) Conduct a comprehensive literature review on the principle and current applications of layer jamming. 2) Develop a finite element model to simulate the behavior of a layer jamming mechanism integrated into the flexible proximal section, focusing on how various parameters influence stiffness. Optionally, perform experimental validation of the developed model to verify its predictive accuracy.

Contact person

Charlotte Deroubaix (charlotte.deroubaix@ulb.be)

references

[1] Upper digestive endoscopy - For a Healthy Belgium. Accessed January 23, 2025, from: https://www.healthybelgium.be/en/medical-practice-variations/digestive-system/gastrointestinal [2] Kohli D, Baillie J. How endoscopes work. Clinical gastrointestinal endoscopy. 2019. 24–31. [3] McCafferty CE, Aghajani MJ, Abi-Hanna D, et al. An update on gastrointestinal endoscopy-associated infections and their contributing factors. Ann Clin Microbiol Antimicrob. 2018;17:36. [4] U.S. Food and Drug Administration. The FDA is recommending transition to duodenoscopes with innovative designs to enhance safety. FDA Safety Communication. 2019 Aug 29. [5] Namburar S, von Renteln D, Damianos J, Bradish L, Barrett J, Aguilera-Fish A, et al. Estimating the environmental impact of disposable endoscopic equipment and endoscopes. Gut. 2022;71(7):1326–1331. [6] Gayam S. Environmental impact of endoscopy: 'Scope' of the problem. Am J Gastroenterol. 2020;15(2):1931–1932. [7] Agrawal D, Tang Z. Sustainability of single-use endoscopes. Tech Innov Gastrointest Endosc. 2021;23:353–362. [8] De Greef A, Delchambre A. Towards flexible medical instruments: Review of flexible fluidic actuators. Precision Eng. 2009;33:311–21. [9] Gifari MW, Naghibi H, Stramigioli S, Abayazid M. A review on recent advances in soft surgical robots for endoscopic applications. Int J Med Robotics. 2019;15(5):e2010. [10] Decroly G, Lambert P, Delchambre A. A soft pneumatic two-degree-of-freedom actuator for endoscopy. Front Robotics AI. 2021;8:768236.

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Vicky Loulas (PhD student)

Research Unit : BEAMS/BIOMECHATRONICS

Description

A closed loop neurostimulator to treat epilepsy

Context

Epilepsy is the second most common chronic neurological disease, associated with stigma and high economic costs. Worldwide, 50 million people are affected by epilepsy, and one-third do not respond to antiepileptic drugs [1]. These patients should be referred for a presurgical evaluation to identify and subsequently remove the epileptogenic focus surgically. If surgery is impossible, neuromodulation can be offered as an adjunctive treatment [2]. In particular, Vagus Nerve Stimulation (VNS) is an attractive neuromodulation technique, as it is less invasive and/or more convenient than other alternatives, i.e., responsive neurostimulation (stimulation is applied directly to the seizure focus), deep brain stimulation of the anterior nucleus of the thalamus, and transcranial direct current stimulation. VNS consists of an implanted pulse generator that delivers trains of electrical pulses to the left vagus nerve, which induces antiepileptic effects for both focal and generalized seizures [3]. Up to 6-9% of patients are rendered seizure-free [4], [5], and approximately half of the treated patients achieve a good clinical response (>50% seizure frequency reduction) [5]. However, despite 30 years of experience in using VNS for epilepsy, the mechanisms of action of VNS remain to be fully elucidated [6]. Nearly one-third of patients do not respond to VNS, and very little is known about why this occurs [4]. Moreover, until now, the titration of VNS parameters is performed empirically, with current intensities raised until the patient’s tolerance or a clinical effect is reached. It may lead to administering unnecessarily high currents, resulting in avoidable side effects and a waste of battery energy [4].

The abortive effect of VNS is confirmed by several human and animal studies [7]–[9]. These publications strengthen the expectation that an automated seizure detection controlling on-demand VNS would significantly increase the treatment's efficiency and provide a warning possibility. Within this context, the vagus nerve is a key bidirectional information pathway between the brain and different visceral organs. For this reason, exploiting the vagus nerve traffic related to seizures might offer a novel method for the early detection of seizures as needed to control an on-demand therapeutic stimulation of the same nerve.

We implemented and validated a chronic recording setup, including specific microcuff electrodes [10], [11]. Our recording systems allow free motion and real-time physiological data acquisition and transfer (including the vagus nerve activity - VENG, electroencephalography - EEG, the video, etc.). It contains a Raspberry Pi, which captures the physiological signal. To do a chronic recording, the user uses a software interface to control the Raspberry Pi from an external server. The setup does not, however, include the stimulation module, which is the missing piece of the puzzle to reach a closed-loop stimulation therapy.

From a technical point of view, closed-loop stimulation is challenging. When the nerve is activated by an electrical stimulus and electrophysiologic activity is recorded by electrodes, the stimulus artifact contaminates the recording. This can be problematic for the design of an implantable device able to simultaneously monitor and stimulate the nervous system (i.e., working in a closed loop). The artifact has several causes, including the voltage gradient induced between the recording electrodes due to the current flowing around the stimulation site, the impedance imbalance between the recording electrodes, and the capacitive coupling between the stimulating and recording leads.

Artifact suppression in neural recording systems is critical, particularly for low-voltage signals such as electroneurogram, where even minimal artifacts from the stimulation can negatively impact the recorded signals.

Objective

This project aims to adapt the chronic recording setup, adding a neural stimulator, to reach a closed-loop stimulation therapy. Major steps will include: - Understanding the chronic recording setup - Designing and implementing a stimulation device - Adding an artifact suppression system to record and stimulate simultaneously from the same electrode - Validating the design on a phantom and then in real conditions.

Students have at their disposal all the project outputs from previous teams.

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be)

references

[1] M. J. Brodie, S. J. E. Barry, G. A. Bamagous, J. D. Norrie, and P. Kwan, “Patterns of treatment response in newly diagnosed epilepsy,” Neurology, vol. 78, no. 20, pp. 1548–1554, May 2012, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/22573629/. [2] D. San-juan, D. O. Dávila-Rodríguez, C. R. Jiménez, M. S. González, S. M. Carranza, J. R. Hernández Mendoza, and D. J. Anschel, “Neuromodulation techniques for status epilepticus: A review,” Brain Stimulation, vol. 12, no. 4. pp. 835–844, 2019. [3] D. M. Woodbury and J. W. Woodbury, “Effects of vagal stimulation on experimentally induced seizures in rats.,” Epilepsia, vol. 31 Suppl 2, pp. S7-19, 1990. [4] D. Labar, “Vagus nerve stimulation for 1 year in 269 patients on unchanged antiepileptic drugs.,” Seizure, vol. 13, no. 6, pp. 392–398, Sep. 2004. [5] D. J. Englot, J. D. Rolston, C. W. Wright, K. H. Hassnain, and E. F. Chang, “Rates and Predictors of Seizure Freedom with Vagus Nerve Stimulation for Intractable Epilepsy,” Neurosurgery, vol. 79, no. 3, pp. 345–353, Sep. 2016, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/26645965/. [6] S. E. Krahl and K. B. Clark, “Vagus nerve stimulation for epilepsy: A review of central mechanisms,” Surg. Neurol. Int., vol. 3, no. SUPPL4, Oct. 2012, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/23230530/. [7] P. Boon, K. Vonck, P. Van Walleghem, M. D’Havé, L. Goossens, T. Vandekerckhove, J. Caemaert, and J. De Reuck, “Programmed and magnet-induced vagus nerve stimulation for refractory epilepsy,” Journal of Clinical Neurophysiology, vol. 18, no. 5. Lippincott Williams and Wilkins, pp. 402–407, 2001, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/11709644/. [8] R. S. McLachlan, “Suppression of Interictal Spikes and Seizures by Stimulation of the Vagus Nerve,” Epilepsia, vol. 34, no. 5, pp. 918–923, 1993, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/8404747/. [9] J. W. Woodbury and D. M. Woodbury, “Vagal Stimulation Reduces the Severity of Maximal Electroshock Seizures in Intact Rats: Use of a Cuff Electrode for Stimulating and Recording,” Pacing and Clinical Electrophysiology, vol. 14, no. 1. Pacing Clin Electrophysiol, pp. 94–107, 1991, Accessed: Oct. 14, 2020. [Online]. Available: https://pubmed.ncbi.nlm.nih.gov/1705342/. [10] J. C. Cerda, E. A. Reina, L. Stumpp, R. Raffoul, L. Vande Perre, M. Diaz Cortes, P. Doguet, J. Delbeke, R. El Tahry, and A. Nonclercq, “Micro Cuff Electrode Manufacture for Vagus Nerve Monitoring in Rats,” BioCAS 2022 - IEEE Biomed. Circuits Syst. Conf. Intell. Biomed. Syst. a Better Futur. Proc., pp. 434–438, 2022. [11] J. Chavez Cerda, E. Acedo Reina, H. Smets, M. Verstraeten, L. Vande Perre, M. Diaz Cortes, P. Doguet, J. Delbeke, R. El Tahry, and A. Nonclercq, “Chronic Setup System for Continuous Monitoring of Epileptic Rats,” in BioCAS 2022 - IEEE Biomedical Circuits and Systems Conference: Intelligent Biomedical Systems for a Better Future, Proceedings, 2022, pp. 591–594.

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Pascal Doguet, Simon Den Haene

Research Unit : BEAMS/BIOMECHATRONICS

Description

Design of a Breathing Synchronizer to address Respiratory Distress Syndrome in Premature Babies

Context

Preterm infants, especially those born before 34 weeks of gestation, are at high risk of developing Respiratory Distress Syndrome (RDS). RDS occurs due to insufficient surfactant, leading to difficulty in breathing and inadequate oxygenation of the blood. Surfactant is a substance that is critical for the proper function of the lungs, particularly in newborns. It is a mixture of lipids and proteins produced naturally in the lungs and works to lower surface tension within the alveoli, the tiny air sacs where gas exchange occurs. By reducing surface tension, surfactant prevents the alveoli from collapsing and allows them to remain open during exhalation, facilitating efficient gas exchange and making breathing easier.

Without enough surfactant, the alveoli tend to collapse after each breath, making it extremely hard for the infant to keep their lungs inflated. Administering surfactant helps keep the alveoli open, allowing for more efficient and less laborious breathing. The administration can be done soon after birth (prophylactic use) or when the symptoms of RDS become apparent (rescue therapy).

The classical technique to administer surfactant involves the insertion of an endotracheal tube by means of laryngoscope, followed by the insertion of a smaller catheter with the surfactant inside the tube. This technique called INSURE, which stands for INtubation, SURfactant administration, and Extubation, is invasive, requires sedation due to the discomfort and also brings risks of physical trauma to the airway such as laryngeal and tracheal injury, vocal cord damage, bleeding or rapid extubation failure [2]. Other techniques were developed to overcome these drawbacks [3], [4], [5], [6]: the LISA technique still uses a laryngoscope followed by administration with a dedicated catheter but can still be a source of pain, especially when performed without any anesthesia. A newly developed technique, called FAST [7] uses a less invasive approach by inserting a slim endoscope, through the nose and delivering the surfactant at the laryngeal level by means of a sub millimeter catheter inserted through the working channel of the endoscope.

The FAST technique also brings the advantage of being able to keep the baby on CPAP (continuous positive airway pressure) while administering the surfactant.

Another approach would be to use nebulization. Surfactant delivery by nebulization would be even less invasive. Some tries have already been done [2][8], delivering nebulized surfactant at the CPAP mask level but surfactant delivery in the lungs is not effective, with a few percents of the volume reaching the target. Therefore, there is a need to go further and enhance the technique.

Objective

This master thesis aims to study and design the breathing synchronizer part of a nebulization catheter with the following structure:

This nebulization catheter is designed to be inserted in the pharyngeal area without the need of endoscopy nor sedation. Nebulization is performed at the tip of the catheter by mixing compressed air and surfactant. As the catheter does not enter the trachea (entry going to the lungs), there is a need to synchronize the nebulization with the inspiration in order to enhance the amount of surfactant going to the lungs and optimize its delivery.

The main goal of this thesis is then to study and design the breathing synchronizer which can be coupled to the nebulization device. It will use a pressure or temperature sensor, either placed in the catheter or external to it (in the mask). The type of sensor and adequateness first needs to be reviewed and defined. Then, using the sensor, the student will develop an electronic board and associated embedded software: to amplify and filter the sensor signal, treat it, detect the breathing pattern and provide a digital control signal to the nebulizer, indicative of inspiration.

This work can also involve the overall management of the nebulization device, which is the goal of another MFE. In this case, it would require some collaboration with the other student.

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be), Pascal Doguet (pascal.doguet@ulb.be), Simon Den Haene (simon.den.haene@ulb.be)

references

[1] Onybiotech, “Onybiotech - what is RDS ?” [2] D. G. Sweet et al., “European Consensus Guidelines on the Management of Respiratory Distress Syndrome - 2019 Update,” Neonatology, vol. 115, no. 4. S. Karger AG, pp. 432–450, Jun. 01, 2019. doi: 10.1159/000499361. [3] M. K. Sabzehei, B. Basiri, M. Shokouhi, S. Ghahremani, and A. Moradi, “Comparison of minimally invasive surfactant therapy with intubation surfactant administration and extubation for treating preterm infants with respiratory distress syndrome: a randomized clinical trial,” Clin Exp Pediatr, vol. 65, no. 4, pp. 188–193, Apr. 2022, doi: 10.3345/cep.2021.00297. [4] G. H. Shim, “Update of minimally invasive surfactant therapy,” Korean J Pediatr, vol. 60, no. 9, pp. 273–281, Sep. 2017, doi: 10.3345/kjp.2017.60.9.273. [5] “Verder et alSurfactant Therapy & Nasal CPAP for Newborns with Respiratory Distress Syndrome1994”. [6] E. Herting, C. Härtel, and W. Göpel, “Less invasive surfactant administration: best practices and unanswered questions,” Current Opinion in Pediatrics, vol. 32, no. 2. Lippincott Williams and Wilkins, pp. 228–234, Apr. 01, 2020. doi: 10.1097/MOP.0000000000000878. [7] D. Guevorkian et al., “Pilot Study to Evaluate a New Method for Endotracheal Administration of Surfactant in Neonatal Respiratory Distress Syndrome: Fiberscope Assisted Surfactant Therapy (FAST),” Neonatology, vol. 119, no. 6, pp. 753–759, Dec. 2022, doi: 10.1159/000525848. [8] J. J. Pillow and S. Minocchieri, “Innovation in surfactant therapy II: Surfactant administration by aerosolization,” Neonatology, vol. 101, no. 4. pp. 337–344, Jun. 2012. doi: 10.1159/000337354. [9] A. Tronde, G. Baran, S. Eirefelt, H. Lennernäs, and U. H. Bengtsson, “Miniaturized Nebulization Catheters: A New Approach for Delivery of Defined Aerosol Doses to the Rat Lung,” 2002. [10] “US20090107503A1_Nebulizing catheter system and methods of use and manufacture”

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Pascal Doguet

Research Unit : BEAMS/BIOMECHATRONICS

Description

Design of an Implantable Cuff Electrode using Laser Induced Graphene

Context

Neurostimulation is the artificial stimulation of brain tissue, peripheral nerves, or muscles to treat neurological disorders or restore impaired or lost body functions. The field of applications is quite large. Here are a couple of examples: deep brain stimulation for Parkinson's disease, cochlear implants providing a sense of sound to people who are deaf, vagus nerve stimulation for the treatment of epilepsy, muscular and sensory nerve stimulation for sensory-motor feedback and restoration for people with paralyzed or lost limbs, to name a few. Neurostimulation is either performed with an external or an implantable neurostimulator. When implanted, it usually consists of an electrical pulse generator, holding the electronics, connected to a lead towards an electrode at the stimulation site.

This master thesis will focus on the electrode part, specifically the design of a cuff electrode, meant to be wrapped around a peripheral nerve. A cuff electrode is composed of the cuff (the flexible material that curls on the nerve), electrically conductive contacts where stimulation occurs, and conductive tracks connecting the contacts to the lead and further to the pulse generator. The main advantage of such an electrode is its self-sizing ability to adapt to the diameter of the nerve [[2], [3]].

Objective

Manufacturing a cuff electrode is a long and tedious process, requiring accurate and intensive craftsmanship. The cuff itself consists of two or more layers of silicone glued together, one of them being pre-stretched before gluing to provide a natural curling ability. Contacts and tracks are embedded on and into the support, requiring welding and/or crimping. The overall process is complex [5], meaning that the price for such an electrode remains expensive and that they are prone to reliability issues and failures.

This master thesis aims to design and manufacture a cuff electrode using a different approach, with the potential advantages of manufacturing simplification and improved electrode characteristics (impedance, reliability, softness, less corrosion). The capabilities of the manufactured electrode will be assessed on a test bench and potentially on small animals.

The proposed manufacturing process and consists of: • A. Realizing a cuff support made of two silicone sheets, one being pre-stretched; • B. Using a sheet of polyimide material (such as Kapton ©) where the electrode contacts and tracks will be structured through a method called laser-induced graphene (LIG) [6]: using a 2D motorized benchtop laser, the polyimide treated by the laser will turn into graphene, an allotrope of carbon with exceptional conductivity and impedance properties; • C. Gluing the resulting sheet of polyimide on the cuff support and isolating tracks.

The main focus of the work evolves around the development of the LIG process, with careful state-of-the-art review, laser and material selection, implementation, and assessment.

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be), Pascal Doguet (pascal.doguet@ulb.be)

references

[1] Saint Luke’s, “Neurostimulator for Vagus Nerve Stimulation.” Accessed: Apr. 15, 2024. [Online]. Available: https://www.saintlukeskc.org/health-library/understanding-vagus-nerve-stimulator-vns-placement [2] G. G. Naples, J. T. Mortimer, A. Scheiner, and J. D. Sweeney, “A Spiral Nerve Cuff Electrode for Peripheral Nerve Stimulation,” 1988. [3] K. A. Yildiz, A. Y. Shin, and K. R. Kaufman, “Interfaces with the peripheral nervous system for the control of a neuroprosthetic limb: A review,” Journal of NeuroEngineering and Rehabilitation, vol. 17, no. 1. BioMed Central Ltd., Mar. 10, 2020. doi: 10.1186/s12984-020-00667-5. [4] M. Ardiem, “Ardiem Medical - Cuff Electrode.” Accessed: Apr. 15, 2024. [Online]. Available: http://www.ardiemmedical.com/neural-cuff/ [5] J. C. Cerda et al., “Micro Cuff Electrode Manufacture for Vagus Nerve Monitoring in Rats,” in 2022 IEEE Biomedical Circuits and Systems Conference (BioCAS), IEEE, Oct. 2022, pp. 434–438. doi: 10.1109/BioCAS54905.2022.9948638. [6] H. Yu, M. Gai, L. Liu, F. Chen, J. Bian, and Y. Huang, “Laser-induced direct graphene patterning: from formation mechanism to flexible applications,” Soft Science, vol. 3, no. 1. OAE Publishing Inc., Mar. 01, 2023. doi: 10.20517/ss.2022.26.

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Nicolas Gaspard (Head of Erasme Neurology Department), Lise Cottin

Research Unit : BEAMS/BIOMECHATRONICS

Description

Graph neural networks to analyze brain connectivity in neurological disorders

Context

Epilepsy is one of the most common collection of neurological disorders, affecting nearly 1% of the global population[1]. It is characterized by an enduring risk of spontaneous seizures and profoundly impacts patients’ lives due to its neurobiological, cognitive, psychological, and social consequences[2]. This work addresses two specific epileptic pathologies that have rarely been explored. The first disorder pertains to patients admitted to an Intensive Care Unit (ICU), where up to 20% will experience nonconvulsive seizures, either because of the acute decompensation of a chronic seizure disorder or as a complication of an acute brain injury or systemic illness[3], [4]. The second concerns children with Spike-Wave Activation during Sleep (SWAS), a pediatric Epileptic Encephalopathy (EE) characterized by massive cognitive-behavioral regression[5]. This condition is one extreme of the spectrum of Self-Limited Focal Epilepsy with Centro-Temporal Spikes (SeLECTS), which is the most common epilepsy syndrome in the pediatric population. In both cases, classical diagnostic approaches based on electroencephalography (EEG) are hampered by several limitations that delay and complicate treatment, which has a highly detrimental impact on patient care[6], [7].

Brain connectivity recently emerged as a promising approach to diagnose neurological disorders, alleviating the shortcomings of traditional EEG analysis. Brain connectivity represents the systemic operation of brain activity via a network depicting the interaction between anatomical areas. Its suitability to study neurological disorders relies on their shared functioning as dysconnectivity syndromes, originating from a reorganization of the neural activity[8]. In particular, it was shown that brain connectivity significantly changes between ICU patients at risk of seizures and controls[9] and from EE-SWAS to typical SeLECTS[10].

Satisfactory performances in diagnosing the pathological cases in the two patient populations have been achieved using logistic regression models fed by connectivity metrics. However, using machine learning models could further improve classification accuracy and thus provide even greater support for diagnosis. This was investigated in previous work at BEAMS, where several machine learning models leveraging functional connectivity were implemented and evaluated on the ICU patient cohort. The models considered explored varying levels of complexity and input formats; from the SVM and Random Forest, fed with connectivity features extracted from the graphs, to the Graph Neural Networks (GNN), taking as input de connectivity matrices directly. Strikingly, it was shown that model performances consistently decreased with complexity.

This preliminary work shows the potential of machine learning methods to identify ICU patients at risk of seizures and prompts to take the analyses further to gain better accuracy and extend the investigation to the SeLECTS cohort. Especially, the performance of the GNNs could be significantly improved, which would give confidence in a completely different way of analyzing connectivity and pave the way for a whole new area of research. Indeed, GNNs preserve the topological properties of graphs, unlike traditional neural networks, which ignore the structural information inherently encoded in graph representations[11].

Objective

This work aims to investigate and assess machine-learning methods based on brain connectivity to predict subsequent seizures in ICU patients and cognitive regression in SeLECTS. The goal would be to compare the performance of machine learning models to the traditional feature-based classification methods. Recording of EEG signals from patients in the ICU and the SeLECTS children will be provided (retrospective data already available).

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be), Lise Cottin (lise.cottin@ulb.be).

references

1. World Health Organization. Epilepsy. https://www.who.int/health-topics/epilepsy#tab=tab_1.
2. Fisher, R. S. et al. ILAE Official Report: A practical clinical definition of epilepsy. Epilepsia 55, 475–482 (2014).
3. Holla, S. K., Krishnamurthy, P. V., Subramaniam, T., Dhakar, M. B. & Struck, A. F. Electrographic Seizures in the Critically Ill. Neurol. Clin. 40, 907–925 (2022).
4. Ruiz, A. R. et al. Association of periodic and rhythmic electroencephalographic patterns with seizures in critically ill patients. JAMA Neurol. 74, 181–188 (2017).
5. Specchio, N. et al. International League Against Epilepsy classification and definition of epilepsy syndromes with onset in childhood: Position paper by the ILAE Task Force on Nosology and Definitions. Epilepsia 63, 1398–1442 (2022).
6. Sánchez Fernández, I. et al. Time to electroencephalography is independently associated with outcome in critically ill neonates and children. Epilepsia 58, 420–428 (2017).
7. Nonclercq, A. et al. Spike detection algorithm automatically adapted to individual patients applied to spike and wave percentage quantification. Neurophysiol. Clin. 39, 123–131 (2009).
8. Catani, M. & Ffytche, D. H. The rises and falls of disconnection syndromes. Brain 128, 2224–2239 (2005).
9. Rubin, D. B. et al. Electrographic predictors of successful weaning from anaesthetics in refractory status epilepticus. Brain 143, 1143–1157 (2020).
10. Bear, J. J., Chapman, K. E. & Tregellas, J. R. The epileptic network and cognition: What functional connectivity is teaching us about the childhood epilepsies. Epilepsia 60, 1491–1507 (2019).
11. Bessadok, A., Mahjoub, M. A. & Rekik, I. Graph Neural Networks in Network Neuroscience. (2021).

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Maxime Verstraeten

Research Unit : BEAMS/BIOMECHATRONICS

Description

Design of an integrated biosensor for automated postoperative monitoring

Context

Postoperative care is critical in preventing severe complications such as respiratory distress, haemorrhage, and infections. Current monitoring relies on intermittent manual measurements by healthcare personnel, which can lead to: • Delayed anomaly detection due to sporadic checks. • Increased workload for medical staff, contributing to fatigue and human error. • Incomplete data integration, as existing devices measure individual parameters (e.g., heart rate, blood pressure, SpO₂) without centralized analysis.

While wearable biosensors exist, most lack: • Multiparametric integration (combining multiple vital signs into a unified system). • Real-time predictive analytics using artificial intelligence (AI) or regular algorithms. • Seamless clinical workflow integration (automated alerts, long-term wearability, minimal patient discomfort).

This project aims to bridge these gaps by developing an autonomous, wireless biosensor that continuously monitors key postoperative parameters and employs sensor fusion and AI/algorithms for early risk detection.

The goal is, therefore, to design a unique biosensor capable of integrating all key parameters of postoperative monitoring and automating their real-time collection. This project, at the intersection of embedded systems, AI/algorithms, and clinical medicine, represents a unique opportunity for future engineers to explore the rapidly growing fields of biomedicine and embedded systems engineering. It is a technical and scientific challenge that directly impacts improving healthcare.

Objective

a) Hardware Development: • Design a wearable, multiparametric biosensor capable of measuring: o Heart rate (via photoplethysmography, PPG) o Blood pressure (non-invasively) o Oxygen saturation (SpO₂) o Respiratory rate o Skin temperature o … ECG and other optional measurements • Ensure low-power operation for prolonged monitoring (24+ hours). • Optimize wearability and patient comfort (wireless connectivity, flexible materials, lightweight wearability-oriented design). b) Software and Data Processing: • Develop real-time sensor fusion algorithms to consolidate data into a unified health indicator, capable of triggering alerts for the medical team. • Implement AI-based anomaly detection (e.g., machine learning for early warning of complications). • Compare AI performance with classical signal processing methods (e.g., threshold-based alerts). c) Validation and Clinical Relevance: • Bench testing: Verify sensor accuracy against medical-grade devices. • Human trials: Initial testing on the student (self-experimentation), followed by controlled clinical validation (if feasible, in collaboration with physicians). • Medical feedback: Collaborate with physicians to assess parameter relevance and usability in real postoperative care. The work thus includes these main steps: • Literature review of the existing sensors, postoperative complications and the relevant monitoring systems. • Sensor design (modalities selection, breadboard and PCB design, …). • Firmware and connectivity (data acquisition, wireless connectivity (BLE, …) with a focus on low-power operation. • Data fusion, algorithms and potentially AI use. • Validation (test-bench of the prototype versus reference devices, and human application). • Clinical feedback and iteration.

Contact person

For more information please contact : Maxime Verstraeten (Teaching assistant, maxime.verstraeten@ulb.be)

references

Bradke, B.; Everman, B. Investigation of Photoplethysmography Behind the Ear for Pulse Oximetry in Hypoxic Conditions with a Novel Device (SPYDR). Biosensors 2020, 10, 34. https://doi.org/10.3390/bios10040034 Choi, Ki Hwan, et al. "In-Ear SpO2: A Tool for Wearable, Unobtrusive Monitoring of Core Blood Oxygen Saturation." Sensors 20, no. 17 (2020). Cleveland Clinic. 2023. ‘Vital Signs: How to Check My Vitals at Home’. Cleveland Clinic. 15 March 2023. https://my.clevelandclinic.org/health/articles/10881-vital-signs Holland, Kimberly. 2023. ‘Blood Oxygen Levels: What Is a Normal Level?’ Healthline. 29 June 2023. https://www.healthline.com/health/normal-blood-oxygen-level iRhythm Technologies. "How Zio Works." iRhythm Technologies. Accessed May 5, 2025. https://www.irhythmtech.com/us/en/patients/how-zio-works Ryals S, Chang A, Schutte-Rodin S, et al. Photoplethysmography–new applications for an old technology: a sleep technology review. J Clin Sleep Med. 2023;19(1):189–195. https://doi.org/10.5664/jcsm.10300.

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Indrani Marchal

Research Unit : BEAMS/BIOMECHATRONICS

Description

Generation of virtual patients for smart wound care electronic patches

Context

The management of chronic wounds is a major clinical and economic challenge. These wounds do not follow the normal healing process and do not reach complete healing after 4 to 8 weeks, often due to stagnation in the inflammatory phase. They can lead to prolonged hospitalisation, additional treatment, a significant reduction in quality of life and an increased risk of serious infections. Their prevalence is estimated at between 1% and 2% in high-income countries, and is tending to increase with the ageing of the population and the rise in co-morbidities such as diabetes and obesity.

Despite advances in treatment, evaluation of the condition of a wound still relies mainly on visual inspection, which is subjective and not very sensitive to early signs of stagnation or infection. Additional tests (cultures, ultrasound, MRI) are costly, sometimes invasive, and do not always provide results within a clinically useful timeframe.

The AI-SWEEP research project aims to design an intelligent dressing integrating sensors for continuous measurement of physiological parameters (temperature, pH, humidity, pressure, proteases, etc.), combined with on-board electronics for data collection and transmission. Eventually, these systems will make it possible to develop AI algorithms to predict healing time or detect infection at an early stage, thereby facilitating remote, personalised care. However, training and validating these algorithms to analyse the data requires large, well-characterised databases, which are difficult to access in a clinical context. To overcome this problem, the aim of this thesis is to generate virtual patients suffering from chronic wounds.

Objective

This master's thesis proposes to generate realistic data from virtual patients suffering from chronic wounds, i.e. simulated time series of biomarkers that sensors integrated into an intelligent bandage could record. This data will be used to reproduce realistic trajectories for different clinical cases (healing, stagnation, aggravation), using modelling based on the scientific literature and an existing database. This will allow various clinical scenarios to be detected and interpreted, and the performance and robustness of prediction algorithms to be tested.

The main steps are as follows: - Carry out a literature review on key biomarkers and their evolution in different clinical scenarios. - Define virtual patient profiles. - Model time series of biomarkers for different clinical cases (healing, stagnation, infection). - Automate data generation. - Visualise, document and validate the results obtained.

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be), Indrani Marchal (indrani.marchal@ulb.be)

attached pdf document

Promotor, co-promotor, advisor : jerome.dohet-eraly@ulb.be, Frank Dubois,

Research Unit : BEAMS/BIOMECHATRONICS

Description

L’analyse in vitro des éléments figurés du sang en flux, notamment les globules rouges et les plaquettes, est essentielle dans l’étude des maladies cardio-vasculaires. Les conditions expérimentales doivent reproduire au mieux les conditions in vivo, ce qui impose d’utiliser des échantillons de plusieurs dizaines de microns d’épaisseur, bien au-delà de la profondeur de champ en microscopie optique classique au grossissement nécessaire. L’analyse dynamique sur les épaisseurs requises est cependant possible en utilisant la microscopie par holographie numérique (MHN), une technologie particulièrement performante permettant d’observer des échantillons dynamiques jusqu’à cent fois plus épais que la profondeur de champ classique. Néanmoins, la densité de particules admissible est limitée, ce qui impose de fortes dilutions du sang. L’objectif de ce mémoire est le développement d’une méthode, basée sur la microscopie par holographie numérique, permettant l’analyse de flux présentant des densités de particules élevées. Pour ce faire, trois axes principaux seront suivis : (1) l’étude de l’imagerie en MHN de flux denses ; (2) la mise au point d’une méthode permettant d’améliorer la densité admissible et (3) la validation expérimentale, notamment sur des échantillons sanguins. Les développements seront d’un grand intérêt, en particulier pour l’étude du mouvement des cellules sanguines dans les écoulements, essentielle pour la compréhension des maladies cardio-vasculaires, première cause de mortalité dans le monde. La méthode sera également d’intérêt pour des applications dans les domaines chimiques ou agro-alimentaires, en vue de l’analyse des milieux hautement concentrés en particules. Le mémoire couvrira tant les aspects théoriques, algorithmiques qu’expérimentaux.

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Amin Hossein, MEng, PhD (ULB, LPHYS), Amin Hossein

Research Unit : BEAMS/BIOMECHATRONICS & CARDIOLOGY DEPARTMENT, UNIVERSITÉ LIBRE DE BRUXELLES

Description

Estimating Pulse Transit Time and Arterial Stiffness Using Seismocardiography and Ballistocardiography Sensors

Context

The assessment of arterial stiffness is a key indicator of cardiovascular health and a predictor of adverse cardiovascular events. A widely used surrogate for arterial stiffness is Pulse Wave Velocity (PWV), which can be estimated non-invasively using Pulse Transit Time (PTT)—the time it takes for the arterial pulse wave to travel between two sites in the circulatory system.

Recent advances in wearable sensors have enabled the measurement of mechanical signals associated with cardiac activity through techniques such as Seismocardiography (SCG) and Ballistocardiography (BCG). SCG captures local chest wall vibrations using accelerometers, while BCG records the recoil of the entire body due to cardiac ejection forces, typically from a sensor placed under the back or feet. When combined, SCG and BCG can offer a unique way to estimate the timing of central and peripheral cardiac events. This enables the estimation of PTT—and therefore provides a proxy for arterial stiffness—using compact, wearable, or even contactless systems.

At the Erasme Hospital, a unique dataset is already available, including simultaneous SCG and BCG recordings from healthy volunteers undergoing bedrest protocols, as well as cardiac MRI-derived PWV measurements. This provides an exceptional opportunity to correlate mechanical signal-derived PTT with ground-truth arterial stiffness values. Additionally, the student will be able to design and conduct a small pilot protocol to test new hypotheses or sensor configurations.

Objective

• Literature Review: The student will begin by reviewing the state of the art on PTT estimation, arterial stiffness, SCG and BCG signal analysis, and relevant validation methods. Emphasis will be placed on understanding the temporal correspondence between signal landmarks and physiological events. • Signal Processing Development: Preprocess SCG and BCG signals to remove noise and motion artifacts, synchronize signals, and identify relevant fiducial points such as the aortic opening or foot of the pulse wave. • PTT Estimation Algorithm: Design and implement algorithms to estimate PTT from the time delay between SCG and BCG features. Explore different combinations of signal landmarks, filters, and signal fusion strategies. • Validation Against MRI-Derived PWV: Compare estimated PTT values with ground truth PWV measurements obtained from cardiac MRI in the existing dataset. Statistical methods will be used to evaluate correlations and model accuracy. • Pilot Protocol Development and Testing: Design a small-scale experimental protocol to test new ideas, such as alternative sensor placements or additional modalities (e.g., photoplethysmography). The student will be involved in data collection and initial analysis.

Scientific environment

The student will collaborate with an interdisciplinary team of engineers, cardiologists, and researchers at the Erasme Hospital’s Cardiology Department. This vibrant environment offers extensive expertise in cardiac signal processing. The intern will also have access to laboratories located at the Erasme Hospital (LPHYS, Department of Cardiology).

Contact person

For more information please contact : Antoine Nonclercq, professor - antoine.nonclercq@ulb.be

references

O. T. Inan et al., “Ballistocardiography and seismocardiography: a review of recent advances,” IEEE J. Biomed. Health Inform., vol. 19, no. 4, pp. 1414–1427, Jul. 2015, doi: 10.1109/JBHI.2014.2361732. A. Hossein et al., “Accurate Detection of Dobutamine-induced Haemodynamic Changes by Kino-Cardiography: A Randomised Double-Blind Placebo-Controlled Validation Study”, Sci. Rep., vol. 9, no. 1, pp. 1–11, Jul. 2019, doi: 10.1038/s41598-019-46823-3. J. Rabineau et al., “Closed-Loop Multiscale Computational Model of Human Blood Circulation. Applications to Ballistocardiography.”, Front. Physiol., 12:734311, Nov. 2021, doi: 10.3389/fphys.2021.734311

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Amin Hossein, MEng, PhD (ULB, LPHYS), Amin Hossein, MEng, PhD (ULB, LPHYS)

Research Unit : BEAMS/BIOMECHATRONICS & CARDIOLOGY DEPARTMENT, UNIVERSITÉ LIBRE DE BRUXELLES

Description

Development and validation of methods to extract cardiac features from sensors within a commercial smartphone

Context

Nowadays, nearly everyone owns a smartphone—and these devices are increasingly equipped with high-quality sensors that can be used to acquire physiological data non-invasively. Specifically, embedded accelerometers and microphones make it possible to record seismocardiography (SCG) and phonocardiography (PCG) directly from the chest using only a smartphone. This opens the door to accessible, portable, and cost-effective tools for cardiac monitoring outside clinical environments. SCG measures subtle chest vibrations caused by cardiac activity, while PCG captures the acoustic signals of heart sounds. Both modalities provide valuable, complementary information about cardiac mechanical events such as aortic valve opening (AO) and mitral valve closure (MC)—key markers of cardiac function.

Despite the promise of SCG and PCG as smartphone-based diagnostic tools, challenges remain, especially in accurately detecting AO and MC. These challenges stem from signal noise, physiological variability, and the complex interplay of cardiac events. However, combining SCG with PCG can help improve robustness and timing accuracy in event detection.

This type of mobile cardiac monitoring is particularly suited for remote or extreme environments—such as for astronauts aboard the ISS, or patients in isolated locations like the Concordia station in Antarctica, or those who cannot easily access hospitals due to geographic or economic barriers.

This master thesis will focus on developing advanced signal processing algorithms to improve the detection of AO and MC, and extract other cardiac metrics from smartphone-acquired SCG and PCG. The aim is to create a reliable, efficient, and fully mobile solution for clinical and remote applications, ultimately contributing to personalized and preventive cardiovascular care.

Objective

• Literature Review: The student will begin by exploring current methods for SCG and PCG signal analysis, with a focus on AO and MC detection techniques, fusion strategies, and mobile implementations. • Signal Processing Development: Design and implement algorithms to preprocess SCG and PCG signals, aiming to reduce noise and enhance key cardiac features. This may include filtering, segmentation, and normalization techniques. • Feature Engineering and Detection: Extract meaningful features from both SCG and PCG that correspond to the mechanical events of AO and MC. Techniques may involve envelope detection, wavelet transforms, and time-frequency analysis. • Validation: Apply the developed methods to existing datasets, including recordings from healthy individuals and patients with known cardiac conditions, to ensure generalizability. • Performance Evaluation: Quantify the method's accuracy, sensitivity, and specificity, and compare it against existing benchmarks. Special attention will be given to evaluating performance across different population groups and recording conditions.

Contact person

For more information please contact : Antoine Nonclercq, professor - antoine.nonclercq@ulb.be

references

O. T. Inan et al., “Ballistocardiography and seismocardiography: a review of recent advances,” IEEE J. Biomed. Health Inform., vol. 19, no. 4, pp. 1414–1427, Jul. 2015, doi: 10.1109/JBHI.2014.2361732. A. Hossein et al., “Accurate Detection of Dobutamine-induced Haemodynamic Changes by Kino-Cardiography: A Randomised Double-Blind Placebo-Controlled Validation Study”, Sci. Rep., vol. 9, no. 1, pp. 1–11, Jul. 2019, doi: 10.1038/s41598-019-46823-3. J. Rabineau et al., “Closed-Loop Multiscale Computational Model of Human Blood Circulation. Applications to Ballistocardiography.”, Front. Physiol., 12:734311, Nov. 2021, doi: 10.3389/fphys.2021.734311

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Karim Zouaoui Boudjeltia (professor, Faculté de Médecine, ULB),

Research Unit : BEAMS/BIOMECHATRONICS & FACULTÉ DE MÉDECINE, ULB

Description

Biophysics of Red Blood Cells and Platelets

Context

Blood is a complex fluid, composed of various cell types, mainly red blood cells (RBCs) and platelets (PLTs), which are the most abundant. Blood plays a fundamental role in the functioning of living organisms. Studying its behavior from a biophysical perspective is essential to deepen our knowledge, especially in specific clinical contexts such as thrombotic phenomena.

Among the key physiological processes, the transport of PLTs to the vessel wall is of major importance. However, this mechanism remains poorly understood on a physical level, although it has profound implications in many pathophysiological processes, such as thrombosis or hemorrhage. Due to its unique nature as a fluid carrying deformable, electrically charged cells, blood eludes a simple description and requires a multidisciplinary approach.

Experimental studies on blood are often limited by difficult observation techniques, while numerical simulations are hampered by the complexity of the system: the diversity of constituents, the deformability of cells, and the multiple physical interactions between them are important obstacles to overcome. Despite significant advances, many questions remain open, both fundamentally and in terms of medical research.

Objective

In this context, we propose to explore experimentally and numerically several key issues: 1. Cell shape variability: What is the role of intra-individual variability in the shape of RBCs and PLTs, and how does this variability influence their dynamic behaviors? 2. Unconventional PLT transport: We have recently shown that, under certain blood flow conditions, PLTs follow a Lévy flight rather than a classical diffusion process. We wish to further investigate this observation and identify the underlying mechanisms. 3. Cellular interactions: Interactions between RBCs and RBCs, as well as between RBCs and PLTs, remain insufficiently understood. However, they are at the heart of the transport properties of blood cells. Our goal is to rigorously quantify these interactions and to elucidate their nature.

It is this last point that we would like to address first.

RBCs are negatively charged by the presence of particular sugars carried by membrane glycoproteins called sialic acids. We would like to feed numerical models with information concerning the role that these sugars could have on RBC-RBC interactions in a first step.

To do this, it would be necessary to expose RBCs, in flow, to an electric field and study the effect of the field on the deflection of the RBCs. In this way, it will be possible to see the importance of electronegative charges on RBC behavior.

Contact person

For more information please contact : Antoine Nonclercq (professor) – antoine.nonclercq@ulb.be

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, - , Enrique Germany (Institute Of NeuroScience - IONS/UCL)

Research Unit : BEAMS/BIOMECHATRONICS & INSTITUTE OF NEUROSCIENCE - IONS/UCL

Description

Development and validation of a photoplethysmography-based system for heart and respiratory rate monitoring in freely moving rodents

Context

The understanding of physiological biomarkers associated with neurological diseases such as epilepsy requires reliable and minimally invasive monitoring tools that can operate under free-behavior conditions in animal models. In epilepsy research, developing chronic multimodal setups has enabled researchers to acquire real-time electrophysiological and behavioral data in rodents. Photoplethysmography (PPG) has emerged as a promising optical technique capable of extracting cardiovascular parameters such as heart rate (HR) and respiratory rate (RR), and it may provide new insights into autonomic function during seizures and neuromodulation procedures.

Previous work within our research group has resulted in two crucial advancements: (1) the derivation and implementation of algorithms for heart rate (HR) and respiratory rate (RR) estimation from photoplethysmography (PPG) signals in rats, and (2) the design of a wearable PPG collar prototype featuring dedicated analog front-end circuitry for high-fidelity signal acquisition. However, no fully integrated solution currently exists that combines custom-designed hardware and signal processing for real-time monitoring within our chronic recording platform, which considers a complete control system to facilitate reliable HR and RR estimation during noisy recordings (e.g., animal movement, collar positioning variability).

Given the limitations of commercial PPG solutions regarding configurability and integration, there is a clear opportunity to develop a dedicated, validated device tailored to rodent physiology and experimental constraints. Such a system would enhance current research capabilities and support the exploration of autonomic biomarkers in the context of epilepsy.

Objective

This thesis project aims to develop and validate a complete PPG monitoring solution for rats, including custom hardware and embedded signal processing algorithms. The proposed work plan is structured as follows: 3.1 Requirement Specification - Review prior projects; define technical and physiological requirements 3.2 Hardware Development - Finalize 3D-printed collar design; develop analog front-end circuit; PCB layout and assembly 3.3 Signal Processing Design - Adapt and optimize algorithms for HR and RR extraction; implement embedded processing logic 3.4 System Integration - Integrate hardware and software; ensure compatibility with existing recording infrastructure 3.5 Bench and In Vivo Tests - Perform benchtop testing and pilot in vivo recordings; analyze signal quality and stability 3.6 Thesis Documentation - Compile results, write thesis; prepare oral presentation

Contact person

For more information please contact : Prof. Antoine Nonclercq (antoine.nonclercq@ulb.be)

references

1. Zhang, Chi, et al. “Respiratory rate estimation from photoplethysmogram baseline wandering by harmonic analysis and sequential fusion.” Biomedical Signal Processing and Control 100 (2025): 107006.
2. Iqbal, Talha, et al. “Photoplethysmography-based respiratory rate estimation algorithm for health monitoring applications.” Journal of medical and biological engineering 42.2 (2022): 242-252.
3. Stankoski, Simon, et al. “Breathing rate estimation from head-worn photoplethysmography sensor data using machine learning.” Sensors 22.6 (2022): 2079.
4. Kyriacou, Panicos A., and John Allen, eds. Photoplethysmography: Technology, Signal Analysis and Applications. Academic Press, 2021.
5. Lester, Lauren Elizabeth, et al. “Design of Pulse Oximeter Solution for Conscious Rodents.” UF Journal of Undergraduate Research 22 (2020).
6. Chavez Cerda, J., et al. “Chronic Setup System for Continuous Monitoring of Epileptic Rats.” BioCAS 2022.
7. Cerda, J. C., et al. “Micro Cuff Electrode Manufacture for Vagus Nerve Monitoring in Rats.” BioCAS 2022.
8. Krahl, S. E., & Clark, K. B. (2012). Vagus nerve stimulation for epilepsy: a review of central mechanisms. Surgical Neurology International, 3(SUPPL4).

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Alec Aeby - Directeur de service de Neurologie Pédiatrique - Faculté de Médecine Hôpital Universitaire des Enfants (HUDE),

Research Unit : BEAMS/BIOMECHATRONICS & SERVICE DE NEUROLOGIE PÉDIATRIQUE - FACULTÉ DE MÉDECINE HÔPITAL UNIVERSITAIRE DES ENFANTS (HUDE)

Description

Sleep Spindle analysis

Context

Epilepsy is the second most common chronic neurological disease, associated with stigma and high economic costs. Worldwide, 50 million people are affected by epilepsy, and one-third do not respond to antiepileptic drugs [1]. These patients should be referred for a presurgical evaluation to identify and subsequently remove the epileptogenic focus surgically. If surgery is impossible, neuromodulation can be offered as an adjunctive treatment [2]. In particular, Vagus Nerve Stimulation (VNS) is an attractive neuromodulation technique, as it is less invasive and/or more convenient than other alternatives, i.e., responsive neurostimulation (stimulation is applied directly to the seizure focus), deep brain stimulation of the anterior nucleus of the thalamus, and transcranial direct current stimulation. VNS consists of an implanted pulse generator that delivers trains of electrical pulses to the left vagus nerve, which induces antiepileptic effects for both focal and generalized seizures [3]. Up to 6-9% of patients are rendered seizure-free [4], [5], and approximately half of the treated patients achieve a good clinical response (>50% seizure frequency reduction) [5]. However, despite 30 years of experience in using VNS for epilepsy, the mechanisms of action of VNS remain to be fully elucidated [6]. Nearly one-third of patients do not respond to VNS, and very little is known about why this occurs [4]. Moreover, until now, the titration of VNS parameters is performed empirically, with current intensities raised until the patient’s tolerance or a clinical effect is reached. It may lead to administering unnecessarily high currents, resulting in avoidable side effects and a waste of battery energy [4].

The abortive effect of VNS is confirmed by several human and animal studies [7]–[9]. These publications strengthen the expectation that an automated seizure detection controlling on-demand VNS would significantly increase the treatment's efficiency and provide a warning possibility. Within this context, the vagus nerve is a key bidirectional information pathway between the brain and different visceral organs. For this reason, exploiting the vagus nerve traffic related to seizures might offer a novel method for the early detection of seizures as needed to control an on-demand therapeutic stimulation of the same nerve.

The circadian rhythm strongly impacts different types of pathology, such as epilepsy, a common neural disease. The rhythmicity of seizures has been known for years and recent advances in EEG monitoring have shown circadian seizure patterns in some patients. Diurnal seizures cluster during certain times of the day, such as on awakening and late in the afternoon, whereas nocturnal seizures occur primarily at bedtime and in the hours before awakening. Seizures occur in patterns dependent on the pathophysiology of the epileptic condition and one hypothesis is that the circadian clock plays a crucial role in these.

We want to study the impact of epilepsy on the night through EEG analysis and, more specifically, through the occurrence of sleep spindles in the EEG. A spindle is commonly defined as a group of rhythmic waves characterized by progressively increasing, then gradually decreasing amplitude that may be present in low voltage background EEG, superimposed to delta activity, or temporally locked to a vertex sharp wave and to a K complex. Spindles are one of the hallmarks of Non-Rapid Eye Movement (NREM) stage 2 sleep, both in adults and children.

Manual scoring of spindles is time-consuming for recordings that typically show 1,000 spindles. Achieving accurate manual scoring on long-term recordings requires a high level of vigilance, resulting in a highly demanding task that augments the risk of decreased accuracy in the diagnosis, especially for sleep-related studies, for which precise information (such as spindle's amplitude, frequency, and length) is often required. In that regard, we developed a sleep spindle detection algorithm.

Objective

The main goal of this project is to analyze EEG sleep spindles in epileptic patients through automated EEG analysis. The sleep spindle analysis software will be used and adapted according to the needs of this analysis.

Major steps will include: - Get familiar with the existing software and the requirements of the neurological team. - Analyze the EEG sleep spindles in epileptic patients. - Propose an adapted version of the EEG sleep spindles analysis software. - Analyze the impact of epilepsy on the night through EEG analysis and, more specifically, through the occurrence of sleep spindles in the EEG.

Contact person

For more information please contact : Antoine Nonclercq (professor) – antoine.nonclercq@ulb.be

references

[1] M. J. Brodie, S. J. E. Barry, G. A. Bamagous, J. D. Norrie, and P. Kwan, “Patterns of treatment response in newly diagnosed epilepsy,” Neurology, vol. 78, no. 20, pp. 1548–1554, May 2012, doi: 10.1212/WNL.0b013e3182563b19. [2] D. San-juan, D. O. Dávila-Rodríguez, C. R. Jiménez, M. S. González, S. M. Carranza, J. R. Hernández Mendoza, and D. J. Anschel, “Neuromodulation techniques for status epilepticus: A review,” Brain Stimulation, vol. 12, no. 4. pp. 835–844, 2019, doi: 10.1016/j.brs.2019.04.005. [3] D. M. Woodbury and J. W. Woodbury, “Effects of vagal stimulation on experimentally induced seizures in rats.,” Epilepsia, vol. 31 Suppl 2, pp. S7-19, 1990. [4] D. Labar, “Vagus nerve stimulation for 1 year in 269 patients on unchanged antiepileptic drugs.,” Seizure, vol. 13, no. 6, pp. 392–398, Sep. 2004, doi: 10.1016/j.seizure.2003.09.009. [5] D. J. Englot, J. D. Rolston, C. W. Wright, K. H. Hassnain, and E. F. Chang, “Rates and Predictors of Seizure Freedom with Vagus Nerve Stimulation for Intractable Epilepsy,” Neurosurgery, vol. 79, no. 3, pp. 345–353

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Karim Zouaoui Boudjeltia (professor, ULB Faculty of Medicine),

Research Unit : BEAMS/BIOMECHATRONICS & ULB FACULTY OF MEDICINE

Description

Lactate in intensive care - room for a new sensor!

Context

Intra-arterial lactate measurement in Intensive Care Unit (ICU) patients with sepsis is of major interest due to its role as a key biomarker of tissue hypoperfusion (hypoxia) and disease severity. Elevation of lactate is associated with increased mortality, while its progressive clearance is an indicator of a favorable response to treatment. The arterial route offers increased accuracy, a more representative measurement of global hemodynamic status, and allows continuous monitoring, which is particularly useful in a critical setting.

Compared to venous measurement, arterial measurement is less influenced by local processes and better reflects systemic perfusion. It is often feasible due to the presence of arterial lines already in place for intensive care. This tool helps guide hemodynamic resuscitation, detect septic shock early, and monitor the effectiveness of therapeutic interventions such as fluids, vasopressors, or oxygenation adjustments.

However, lactate interpretation must take into account metabolic processes unrelated to hypoperfusion, and its use requires clinical expertise. Although the method is invasive, the advantages in terms of rapid decision-making and personalized care make it a valuable tool for optimizing the management of patients with sepsis and improving their chances of survival.

Objective

Propose a sensor measuring lactate continuously, which would be placed at the end of catheters already used to measure other parameters such as blood pressure in real time.

This work requires a chemical-biochemical approach, microelectronics and signal processing.

Contact person

For more information please contact :Antoine Nonclercq (professor) – antoine.nonclercq@ulb.be

attached pdf document

Promotor, co-promotor, advisor : antoine.nonclercq@ulb.be, Benoit Scheid, Ramzi Ben Hassen, Cyril Tordeur

Research Unit : BEAMS/TIPS/CARDIOLOGY DEPARTMENT

Description

Development and optimization of a pressure myography system based on an open-source design

Context

Pressure myography is an advanced technique used in vascular research to study the physiological properties and functional responses of small arteries, veins, and other blood vessels under near-physiological conditions [1, 2]. This method involves isolating a small segment of a vessel, mounting it onto glass cannulas, and pressurizing it to simulate the transmural pressure experienced in vivo [3]. By maintaining these conditions, pressure myography allows researchers to measure key parameters such as vessel diameter, wall thickness, vascular resistance, and shear stress [4]. The system is particularly valuable for investigating intrinsic vascular responses like myogenic tone (the vessel’s ability to constrict or dilate in response to pressure changes), flow-mediated dilation, and the effects of pharmacological agents on vascular function [5, 6]. Pressure myography is also used to evaluate structural changes in vessel walls and their intrinsic biomechanical properties under varying conditions [7, 8]. Its ability to replicate physiological pressures and flows makes it a superior alternative to traditional wire myography for studying resistance vessels that play a critical role in regulating blood flow and peripheral resistance.

Overall, pressure myography is an essential tool in cardiovascular research, offering insights into the dynamic interplay between vascular structure and function. It has applications ranging from basic physiological studies to drug development and understanding pathophysiological processes in cardiovascular diseases.

Objective

This master thesis aims to manufacture a pressure myography system based on an existing open-source design, the VasoTracker. VasoTracker originated from a need to find a cost-effective alternative to expensive commercial systems used for studying blood vessel function. The initial VasoTracker system was a custom pressure myograph bath coupled with software designed for the live measurement of blood vessel diameter. Recognizing that other researchers might face similar challenges and benefit from our solution, researchers from the University of Strathclyde and the University of Glasgow released these tools as open-source resources.

In addition to the manufacturing process, a design optimization might be warranted to address the impossibility to analyze multiple vessels in parallel. Additionally, further work might be conducted to optimize the available design in order to reach an even more physiological condition. Upon termination of the manufacturing process, the system will be assessed on rodent arteries in a research project investigating the effects of weightlessness on central vascular structure and function. The focus of the work evolves around the manufacturing of an open-source pressure myography system, with assessment, and room for design improvement for further work.

Contact person

For more information please contact : Antoine Nonclercq (antoine.nonclercq@ulb.be), Ramzi Ben Hassen (ramzi.ben.hassen@ulb.be)

references

1. Wenceslau, C.F., et al., Guidelines for the measurement of vascular function and structure in isolated arteries and veins. Am J Physiol Heart Circ Physiol, 2021. 321(1): p. H77-H111.
2. Schjorring, O.L., R. Carlsson, and U. Simonsen, Pressure Myography to Study the Function and Structure of Isolated Small Arteries. Methods Mol Biol, 2015. 1339: p. 277-95.
3. Shahid, M. and E.S. Buys, Assessing murine resistance artery function using pressure myography. J Vis Exp, 2013(76).
4. Lawton, P.F., et al., VasoTracker, a Low-Cost and Open Source Pressure Myograph System for Vascular Physiology. Front Physiol, 2019. 10: p. 99.
5. Hanson Ng, Y.Y., et al., Asymmetric Dimethylarginine Enables Depolarizing Spikes and Vasospasm in Mesenteric and Coronary Resistance Arteries. Hypertension, 2024. 81(4): p. 764-775.
6. Masterova, K.S., et al., Enhancing flow-mediated dilation analysis by optimizing an open-source software with automated edge detection. J Appl Physiol (1985), 2024. 137(2): p. 300-311.
7. Navindaran, K., J.S. Kang, and K. Moon, Techniques for characterizing mechanical properties of soft tissues. J Mech Behav Biomed Mater, 2023. 138: p. 105575.
8. Akhtar, R., et al., Characterizing the elastic properties of tissues. Mater Today (Kidlington), 2011. 14(3): p. 96-105.

attached pdf document

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - , Menthy Denayer, María Alejandra Díaz

Research Unit : BRUBOTICS

Description

Context

Musculoskeletal simulations enable researchers to analyse difficult-to-measure data, like muscle activations, joint contact forces and energetic costs. Therefore, they provide many opportunities to study the effect of assistive devices, like exoskeletons or prostheses. Moreover, these simulations allow us to optimize the exoskeleton’s assistance to the user, to reduce, for example, their energetic cost. As such, musculoskeletal simulations can be a worthwhile alternative to real-life human-in-the loop optimization (HILO) experiments, which are often lengthy and involved.

Objective

The goal of this thesis is to model a (basic) exoskeleton model in the musculoskeletal software OpenSim. Additionally, the student will use optimization, to adapt the exoskeleton assistance to the user in simulation.

The two main research questions will be: 1) to investigate how we can model the exoskeleton assistance inside of OpenSim, and 2) to analyse the effect of optimizing the exoskeleton assistance for the user's model.

Depending on the student’s own interests, there is a possibility to focus on different aspects of the simulation (i.e. optimization, cost functions, modelling, upper-limb, lower-limb etc.).

Methods

The student will use musculoskeletal modelling in OpenSim to model a (basic), one degree of freedom exoskeleton. This will involve familiarizing themselves with the available tools (i.e. scaling, inverse kinematics, inverse dynamics, static optimization) by following the available tutorials. Additionally, they will use the OpenSim API to create an optimization script. The student will also be asked to carefully document their process so that their work can be re-used.

Prerequisites

There are no specific prerequisites for the thesis. Knowledge of MATLAB, or programming in general, is a plus.

Contact person

For more information please contact: menthy.denayer@vub.be or ma.diaz@vub.be.

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - ,

Research Unit : BRUBOTICS

Description

Mid-level control of an active shoulder exoskeleton

Context

Musculoskeletal disorders are a major reason for absenteism in the workplace. Exoskeletons enable to reduce the injury risk by taking over part of the load. Active exoskeletons are more versatile since they can change their support specific to the situation. However, this requires an advanced controller.

Objective

This project will deal with the mid-level control of the exoskeleton, i.e. the translation layer which generates the torque commands for the actuators. Apart from using the traditional encoders and/or IMU sensors, pressure and strain measurements at the interface will be used to enrich the knowledge of the system. There are, however, some practical problems associated with the use of pressure and strain sensors. In this project, you will attempt to solve these problems by combining machine learning and physics-based modelling. The final goal of this project is to have an active shoulder exoskeleton which can provide adequate support for various payloads.

Methods

First, you will acquire an extensive dataset of test subjects lifting different loads with the exoskeleton under different conditions. Using this data, you will map the relationship between the IMU/pressure/strain measurements and the payload. This algorithm will then serve as the basis for a mid-level controller.

Things you will work on

• Acquiring data in a lab setting (i.e., performing tests where subjects lift loads in different conditions)
• Data processing
• Modelling and machine learning
• Control

Contact person

For more information please contact : Tom.Verstraten@vub.be

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - ,

Research Unit : BRUBOTICS

Description

Context

Comfort is a major concern in orthopaedic devices like prosthetic sockets and exoskeletons. These devices must transmit substantial forces to the human limbs without creating excessive pressure concentrations, something which is extremely challenging to achieve. In the field of prosthetics, orthopaedic technicians rely on years of experience to design functional prosthetic sockets. In the field of exoskeletons, it is still poorly understood how appropriate interfaces should be designed. Our team’s goal is to improve or even automate the design of these interfaces by simulating the interaction between the device and the limb. This would require an appropriate model of the limb, both in terms of shape and material properties. However, state-of-the-art models currently do not capture the human limb properties well.

Objective and methods

The objective of the thesis is to develop a finite element model of the human limb. The material properties of the limb will be identified through inverse finite element analysis, using indentation experiments as an input. This method is described in Ballen-Moreno et al. (2023) – see reference below. The student will further develop this method by acquiring indentation data and 3D surface scans, and processing these data in a FEA software package. The student will further explore methods to segment the limb and explore the level of detail needed to acquire sufficiently accurate simulations of the interaction between the limb and the orthopedic device (either a prosthetic socket or an exoskeleton interface).

Prerequisites

• Finite element modelling (Abaqus, Inventor or similar)
• Hands-on mentality – you will be doing tests with people
• Curiosity and critical thinking – You will need to explore and tune different models, which will involve some trial and error.

Contact person

For more information please contact : Tom.Verstraten@vub.be

References

Ballen-Moreno, F., Langlois, K., Ferrentino, P., Brancart, J., Van Vlerken, C., Vanderborght, B., ... & Verstraten, T. (2023). Robotically Aided Method to Characterise the Soft Tissue Interaction with Wearable Robots. In 2023 International Conference on Rehabilitation Robotics (ICORR) (pp. 1-6). Link: https://ieeexplore.ieee.org/abstract/document/10304757

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - , James Flack

Research Unit : BRUBOTICS

Description

Project description

Ankle exoskeletons have a large potential both in rehabilitation and partial assistance in daily life. However, adding mass to the ankle has a detrimental effect on the usability of these devices. Flexible shafts are a relatively unexplored method of remote actuation and are used for transmitting rotary motion between 2 points that are not fixed to one another. The use of a flexible shaft in an ankle exoskeleton introduces challenges and advantages to the design of the mechanical structure and the actuation of the device.

The aim of this thesis would be exploring the design of an ankle exoskeleton driven by a flexible shaft where torque is transmitted form the lower back or hip down to the ankle with a focus on the actuation system or the mechanical design of aspects of the device.

Prerequisite

Experience with mechanics, mechanical design and CAD will be useful for this project.

Contact person

For more information please contact : james.flack@vub.be

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - , James Flack

Research Unit : BRUBOTICS

Description

Project description

Ankle exoskeletons have a large potential both in rehabilitation and partial assistance in daily life. However, due to the ankle being a large distance from the centre of mass, adding mass to the ankle has a detrimental effect on the usability of these devices. Flexible shafts are a relatively unexplored method of remote actuation and are used for transmitting rotary motion between 2 points that are not fixed to one another, in this case from the lower back or hip down to the ankle.

The aim of this thesis would be developing a mid level controller for the ankle in the plantarflexion and dorsiflexion movements. One key area for exploration will be on how the flexible shaft can be sensorized and used to allow for more robust control.

This thesis will involve research on suitable ways to equip the shaft with sensors and identifying controllers that can leverage the additional information available from the flexible shaft. A dataset will need to be developed by the student to allow for development of an offline simulation of the controller. If there is enough time, an online version of the controller could be developed in order to further validate their results.

Prerequisites

Experience with system identification, sensors and controller design will be useful for this project.

Contact person

For more information please contact : james.flack@vub.be

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - , Stijn Hamelryckx, Nathan De Smedt, Louis Flynn

Research Unit : BRUBOTICS

Description

Improving interfaces for exoskeletons – What factors influence efficient torque transmission?

Exoskeletons are assistive devices that can help populations who suffer from gait impairment. They do so by providing assistive torques (or forces) on the wearer. However, due to the soft tissue of the attachment points (needed for comfort) and those of the wearer, the torque transmission from the assistive device to the wearer comes with significant loss due to for example deformation in the attachment. These transmission losses lead to multiple negative impacts: they reduce the actual assistance the wearer receives, the controllability and accuracy of the exoskeleton but also lead to parasitic forces that can cause discomfort to the wearer. Even though solving these problems is crucial for a widespread adaptation of exoskeletons, very little is known about the underlying behavior of the problem and mitigation methods.

In this thesis, the student will try to identify some of the design considerations that could increase efficiency by looking at the stiffness, placement, size, and shape of the cuff and relating these to relative displacements. The student will study these parameters through prototyping, testing, and comparison to find a set of design guidelines that can improve torque transmission (while maintaining comfort) through the cuff for future exoskeletons. Finally the student will come up with a cuff proof of concept that will be tested on available exoskeletons.

Prerequisites:

CAD design

Python

Contact person:

Stijn.hamelryckx@vub.be, Nathan.elie.o.de.smedt@vub.be or Louis.flynn@vub.be

Promotor, co-promotor, advisor : ilias.el.makrini@vub.be, Mohsen Omidi, Mohsen Omidi

Research Unit : BRUBOTICS - MECH - AIXC

Description

Development of a Virtual Reality (VR) Environment for Human-Robot Collaboration in Unity3D

Context

This project aims to design and implement a virtual reality (VR) platform that allows immersive simulation of human-robot collaboration (HRC). The focus is to create a Unity3D-based environment where a user, equipped with a VR headset and motion controllers (e.g., HTC VIVE Pro, Manus Glove), can interact with a simulated robotic arm (e.g., FRANKA/PANDA) to complete collaborative tasks such as object selection, gesture-based commands, or coordinated manipulation.

Objective

The user’s actions (e.g., pointing, hand signals, task demonstrations) are captured and interpreted to guide robot responses. For instance, the robot can identify and pick an object indicated by user gestures, or mimic a demonstrated motion path (e.g., painting). This setup supports safe testing, training, and prototyping of human-robot workflows in virtual environments.

Methods

The environment will be developed in Unity3D using C#, with interactive feedback and usability considerations.

Prerequisite

• Unity3D,
• C# language,

Contact person

For more information please contact : Dr. ir. Mohsen OMIDI E-mail to contact person:

Mohsen.Omidi@vub.be

Number of possible students: 1

attached pdf document

Promotor, co-promotor, advisor : ilias.el.makrini@vub.be, Mohsen Omidi, Mohsen Omidi

Research Unit : BRUBOTICS - MECH - AIXC

Description

Robots and Unity3D in C#: Design and Assessing New Control Mechanisms

This project explores different methods for controlling robots using Unity3D and C#. Unity3D offers a flexible simulation environment, while C# provides powerful tools to implement control strategies. These methods are crucial for various applications like manufacturing, healthcare, and entertainment, where precise and efficient robot control is vital.

Objective

The goal is to implement and compare various control techniques for robots in Unity3D, focusing on: Direct input (keyboard, joystick) Autonomous control (pathfinding, AI) Sensor-based control (virtual sensors) Each method will be analyzed based on precision, responsiveness, and efficiency in simulated real-world scenarios. Key control strategies include: • Manual Input: User control via keyboard or joystick. • Autonomous Control: Robots will navigate using AI and pathfinding algorithms. • Sensor-Based: Robots will respond to virtual sensors like proximity and gyroscopes. These will be developed using Unity’s physics engine and C# scripting.

Methods

The project will provide a comparative analysis of robot control methods, identifying strengths and weaknesses. This will contribute to future research in robotics and automation, demonstrating practical real-world robot behaviors in Unity3D.

Prerequisite

• Unity3D
• C#

Contact person

For more information please contact : Mohsen.Omidi@vub.be

Number of possible students: 1

attached pdf document

Promotor, co-promotor, advisor : ilias.el.makrini@vub.be, - , Tom Turcksin

Research Unit : BRUBOTICS R&MM

Description

Object Detection for Force Estimation in Biomechanical Assessment

The project aims to estimate interaction forces during human-object interaction using vision-based object detection and tracking, improving biomechanical assessments and ergonomic analysis.

Context

The project is done in collaboration with AugmentX (www.augmentx.be). It uses a pair of RGB-D (depth) cameras to capture hand-object interactions during physical tasks.

Objective

To develop a computer vision system that detects and tracks objects in real time and estimates the forces applied by the user’s hands based on object properties such as shape, mass, and acceleration.

Methods

Different methods are to be tested, namely:

*Literature review on object detection, tracking, and force estimation techniques;

*Implementation of a detection and tracking system using RGB-D cameras;

*Development of an algorithm to estimate interaction forces from object motion and known properties;

*(Optional) Extending the system to work with moving cameras;

*Validation against reference force data in controlled experiments.

Prerequisite

• Matlab,
• Python or C++

Contact person

For more information please contact: tom.turcksin@vub.be

Promotor, co-promotor, advisor : ilias.el.makrini@vub.be, - , Ilias El Makrini

Research Unit : BRUBOTICS R&MM

Description

Development of a Modular Simulation for Ergonomic Task Analysis in Unity3D

The project aims to build a proof-of-concept simulation and post-processing pipeline for ergonomic analysis using Unity3D for human motion generation and MATLAB for data processing and visualization.

Context

The aim is to enable rapid prototyping and evaluation of human task performance based on simulated movements, especially in industrial and assembly contexts.

All work will be conducted in simulation (Unity3D), so no lab hardware is needed. Motion data will be generated from known task constraints and processed offline in MATLAB. The developed architecture will serve as a basis for more advanced biomechanics or real-time optimization studies in the future.

Objective

To develop a modular software that simulates human task execution in Unity3D and extracts joint-level data for ergonomic post-processing in MATLAB. The simulation will be used to analyze joint angles and posture scores, forming the foundation for future real-time ergonomic feedback systems.

Methods

The project will consist of several key phases:

• Architecture design: A software architecture will be designed to integrate Unity3D (for simulation) with MATLAB (for analysis). This includes task definition, movement simulation, and data export structures.

• Avatar & task setup in Unity: A human avatar will be imported into Unity3D (e.g., from Mixamo), and a few sample industrial tasks (e.g., object reach, lifting) will be animated using inverse kinematics or keyframe animations.

• Joint data extraction: Scripts will be implemented in Unity to log joint angles and hand trajectories during task execution, exported in a format readable by MATLAB.

• Post-processing in MATLAB: MATLAB scripts will be developed to calculate simple ergonomic scores (e.g., RULA, EAWS approximation) and plot joint usage statistics.

• GUI prototype: A basic interface in Unity or MATLAB that allows users to select which metrics or body segments to visualize.

This project can be carried out by two students working in parallel (e.g., one focusing on Unity simulation, the other on MATLAB post-processing).

Prerequisite

• C#, C++,
• Matlab

Contact person

For more information please contact : ilias.el.makrini@vub.be

Promotor, co-promotor, advisor : fabrizio.pucci@ulb.be, Prof. Marianne Rooman,

Research Unit : COMPUTATIONAL BIOLOGY AND BIOINFORMATICS

Description

The prediction of drug response in cancer cell lines has become a key strategy for identifying molecular factors that influence treatment efficacy. Datasets such as Genomics of Drug Sensitivity in Cancer (GDSC) have provided fundamental information to link genomic and transcriptomic features to drug sensitivity. In this master's thesis, we aim to develop a machine learning framework to predict drug responsiveness by integrating multiple layers of biological information. Specifically, we will extract and engineer features from genomic and transcriptomic data, while also incorporating protein structural information to enhance the biological interpretability of the model. Beyond modeling drug response in cancer cell lines, we will extend our approach to a more clinically relevant scenario by accounting for tumor clonal heterogeneity—a key challenge in personalized oncology. In this phase, we will identify driver mutations from patient-derived genomic cohorts and assess their potential impact on drug sensitivity. This will be achieved by predicting how these mutations alter protein function and, in turn, influence treatment response. The ultimate goal of this work is to move toward a precision oncology framework, where we can computationally predict the most effective drugs or drug combinations for individual patients based on their unique molecular profiles.

For inquiries, please contact: Fabrizio.Pucci@ulb.be

Promotor, co-promotor, advisor : fabrizio.pucci@ulb.be, Prof. Marianne Rooman,

Research Unit : COMPUTATIONAL BIOLOGY AND BIOINFORMATICS

Description

Maintaining global health requires the development of generic and versatile technologies that allow fast and effective responses to the large variety of disorders, in particular cancer and emerging infectious diseases. Among these, monoclonal antibodies (mAbs) play an important role. Indeed, antibodies can bind antigens, such as bacterial or viral proteins or proteins expressed in cancer cells, and trigger the human immune response. In this project, we will build a bioinformatics pipeline for the design of mAbs that can bind with high affinity a specific target antigen. For that purpose, we will rely on experimentally characterized antibody-antigen complexes and their binding affinity values, detect informative sequence- and structure-based features, and combine them into a predictor using artificial intelligence techniques. The mAbs that we will design will be tested in vitro by experimental collaborators. While the project is focused on the design of a generic pipeline, it can be applied to specific case studies such as chronic lymphocytic leukemia, on which we are currently collaborating with cancer immunologists (Institute Bordet, BE), or infectious in collaboration with the CER research center.

Promotor, co-promotor, advisor : gilles.bruylants@ulb.be, Maurice Retout,

Research Unit : EMNS

Description

Photoacoustic Imaging (PAI), the fastest growing biomedical imaging modality in the last decade, has the potential to significantly impact the field of nanomedicine. It is non-ionizing, non-invasive and uses a nanosecond pulsed laser to generate pressure waves that can be detected by conventional ultrasound transducers. Because PAI uses a light-in-sound-out approach, it has the strengths of ultrasound, i.e. good tissue penetration, real-time monitoring, low cost and high spatial resolution, but also the high contrast, specificity and sensitivity of optical methods. Although endogenous contrast agents such as oxygenated or deoxygenated hemoglobin and melanin can be used, PAI still lacks exogenous contrast agents, which could increase sensitivity and allow targeting of specific cells (such as cancer cells). This improved diagnostic capacity could also be combined with therapeutic activity, creating new and promising theragnostic platforms. The EMNS laboratory is involved in the development of such functionalized nanomaterials based on gold nanorods and copper sulfide nanoparticles.

Students will synthesize and characterize the functionalized nanostructures (UV-Vis, IR, DLS, TEM, …) and evaluate their potential as contrast agents for PAI and photothermal transducers.

Contact person For more information please contact : Gilles.Bruylants@ulb.be; Maurice.Retout@ulb.be

Promotor, co-promotor, advisor : gilles.bruylants@ulb.be, Maurice Retout,

Research Unit : EMNS

Description

Delivery of anticancer peptides combined to mild-photothermal therapy for improved cancer treatment.

Context

Today, cancer is one of the leading causes of death. Current treatments involve mostly chimio- and radiotherapies but they lack specificity and cause damages to healthy tissues. Anticancer peptides (ACPs) are short polycationic peptide sequences, typically produced by microorganisms, that have anticancer properties via membrane disruption, pore formation or metabolisms disfunction. The efficiency of ACPs can be enhanced when combined to mild photothermal therapy (PTT) using plasmonic nanomaterials. Mild PTT (<45°C) is safer for healthy tissues than traditional PTT (>60°C). Combining ACPs delivery and mild PTT could thus be a promising strategy for specific and safe cancer treatments. Cancer cells overexpress various biomolecules such as enzymes. This dysfunction in their metabolism can be used for targeted therapy.

Objective

The goal of this project is to use plasmonic nanomaterials (gold nanorods and/or silver nanoplates) decorated with ACPs for cancer treatments. Gold nanorods and silver nanoplates have absorption in the near-infrared region (700 nm-950 nm) and can thus be used as thermal transducer in vivo. We have developed specific surface modifications of these nanomaterials that allow precise control over the conjugation with peptides. Also, we want to use ACPs that are enzyme-responsive to increase the specificity of the peptide release.

Methods

Functionalized gold nanorods and silver nanoplates will be synthesized and then conjugated to ACPs. Various ACPs will be investigated as well as the modification of the nanomaterials with targeting peptides to enhance their selectivity. For every combination of ACPs and targeting peptides, the materials will be characterized using a wide variety of techniques such as UV-Vis, IR, TEM, DLS, … The capacity to release the ACPs form the material surface in the presence of enzyme will be systematically investigated. Promising materials in term of thermal conversion, peptide delivery and colloidal stability will be used in the cytotoxicity study on colorectal cancer and glioblastoma cell lines.

Contact person

For more information please contact :

Promotor, co-promotor, advisor : gilles.bruylants@ulb.be, - ,

Research Unit : EMNS

Description

Design of functionalized nanomaterials for biomedical applications.

Context

The EMNS Laboratory is developing various types of nanomaterials for biomedical applications, including: • Gold nanoparticles for RNA delivery to specific organs • Gold nanorods for photothermal therapy • Iron oxide nanoparticles as MRI contrast agents All these applications benefit from a specialized surface treatment developed in collaboration with the Organic Chemistry Laboratory of ULB (LOC-ULB, Prof. I. Jabin).

Objective

The primary goal of these projects is to enhance the effectiveness of each application by leveraging the unique properties provided by calixarene-based coatings. These properties include exceptional stability, precise control over ligand density, and excellent biocompatibility. To tailor the nanoparticles for specific biomedical uses, an additional coating with biomolecules—such as antibodies, peptide aptamers, or nucleic acids—is required.

Methods

Nanoparticles will first be synthesized and functionalized with a calixarene-based layer. Various chemical strategies will then be explored to attach therapeutic agents and/or targeting ligands at defined densities. The physicochemical and biological properties of the particles will be evaluated in vitro, prior to any in vivo testing. Depending on their background, students may also participate in the in vitro testing phase, including evaluating the performance of these nanodevices in cell culture experiments.

Contact person

For more information please contact : gilles.bruylants@ulb.be

Promotor, co-promotor, advisor : gilles.bruylants@ulb.be, Dr. Maurice Retout,

Research Unit : EMNS

Description

Nucleotide-peptide complex for the specific delivery of miRNA

Context

MicroRNA have emerged has promising candidates for targeted therapy. They can interact with the cell metabolism and restore (or inhibits) the normal (or abnormal) function at the cellular level. The main challenge to deliver miRNA is the specificity of the distribution and the protection of the nucleotide from degradation by endogenous enzymes. Various strategies have been investigated to carry the nucleotide to the cells such as noble metal nanoparticles or lipidic vesicles, but they have either a poor body clearance or a high toxicity, respectively. Developing new strategies are thus highly necessary.

Objective

We have recently developed self-assembled peptide-nucleotide complexes. By using specific peptide sequence, we could produce stable RNA-peptide assembly with interesting optical properties. We want now to demonstrate its potential for miRNA delivery associated to optical imaging modalities. We want to engineer the peptide sequences to ensure appropriate targeting of the RNA delivery, and two applications are currently under investigation: cancer and Alzheimer’s disease.

Methods

Students will investigate the assembly between peptides and nucleotides with various methods such as UV-Vis, emission, DLS, TEM, … Various sequences will be studied to understand the mechanism controlling the assembly. Particularly, isothermal calorimetry, that is an original technique for which the EMNS has a top-notch expertise, will be used to determine the affinity constants and the stoichiometry of the RNA-peptide assembly. When the optimal peptide sequence will be determined, studies on cells and animals will be carried out. The student will use an RNA carrying an infrared fluorophore for the tracking of the RNA delivery. The cell internalization will be investigated with confocal microscopy on cell cultures and the biodistribution will be studied by fluorescence imaging on murine model.

Contact person

For more information please contact : gilles.bruylants@ulb.be or maurice.retout@ulb.be

Promotor, co-promotor, advisor : hennie.valkenier@ulb.be, Benoit Scheid, Hasna Ummat, Majid Layachi

Research Unit : EMNS AN TIPS

Description

Project title

Preparation of Giant Unilamellar Vesicles using microfluidics and their application for transmembrane transport.

Context

In the EMNS laboratory, vesicles prepared from natural lipids (liposomes) are used as models for cell membranes to study processes like transmembrane transport and cell targeting. With standard procedures, vesicles with diameters of up to 200 nm are easily made, but the preparation of giant unilamellar vesicles (GUVs, ~10 µm) is still a challenge. The TIPS laboratory is specialized in microfluidics, which can be used to prepare droplets and vesicles.

Objective

The aim of this collaborative project is to develop a method to prepare GUVs as membrane model system by microfluidics and to study the transport of ions across the lipid membranes of these GUVs

Methods

You will use a home-made 3D-printed micro-emulsion generator to produce double emulsions and screen the conditions (fluid viscosities, lipid solutions and concentration, flow rates, geometry) to identify the optimal regime for generating stable GUVs, with minimal organic solvent present. You will then characterise these GUVs by fluorescence microscopy, using encapsulated (water-soluble) fluorescent probes, and by using fluorescent compounds in the membrane. Changes in the fluorescence will then be used to monitor the transmembrane transport of ions.

Prerequisite

Chemistry and Spectroscopy

Contact person

For more information please contact : hennie.valkenier@ulb.be and benoit.scheid@ulb.be

Promotor, co-promotor, advisor : gilles.bruylants@ulb.be, Maurice Retout,

Research Unit : ENGINEERING OF MOLECULAR NANOSYSTEMS

Description

Development of activable NIR contrast agents for photoacoustic imaging.

Context

Photoacoustic Imaging (PAI), the fastest growing biomedical imaging modality in the last decade, has the potential to significantly impact the field of nanomedicine. It is non-ionizing, non-invasive and uses a nanosecond pulsed laser to generate pressure waves that can be detected by conventional ultrasound transducers. Because PAI uses a light-in-sound-out approach, it has the strengths of ultrasound, i.e. good tissue penetration, real-time monitoring, low cost and high spatial resolution, but also the high contrast, specificity and sensitivity of optical methods. Although endogenous contrast agents such as oxygenated or deoxygenated hemoglobin and melanin can be used, PAI still lacks exogenous contrast agents, which could increase sensitivity and allow targeting of specific cells (such as cancer cells). The EMNS laboratory is involved in the development of such functionalized nanomaterials based on gold nanorods, silver nanoplates and copper sulfide nanoparticles.

Objective

Development of activable contrast agent for photoacoustic imaging. We want to develop contrast agent that produce higher photoacoustic signal upon enzymatic activity.

Methods

Students will synthesize the nanomaterials with various shape, size or surface chemistries and characterize them with several techniques. The nanomaterials will be modified with appropriate peptide sequences to make it enzyme sensitive. The performance in photoacoustic imaging will be systematically investigated both in vitro and in vivo. The activity of the probe upon enzymatic activity will be first investigated with recombinant proteases or kinases. The biodistribution

Contact person

For more information please contact : gilles.bruylants@ulb.be or maurice.retout@ulb.be

Promotor, co-promotor, advisor : hennie.valkenier@ulb.be, - , Karolis Norvaisa

Research Unit : ENGINEERING OF MOLECULAR NANOSYSTEMS

Description

Project title

Improving the transport of anions and negatively charged drugs into cells.

Context

Transport of ions across membranes is an important process in biology. Specialized proteins embedded within the cellular membranes take care of this. In our laboratory, we seek to mimic the action of these proteins with synthetic molecules that can transport anions across membranes.

Objective

Many important molecules in biology have anionic phosphate or carboxylate groups. The aim of this project is to study how different anionic molecules, especially drugs or models for drugs, can cross the membrane and how this process can be enhanced using synthetic anion receptors.

Methods

This will involve the preparation of liposomes, spherical assemblies of lipids, as models for cell membranes. You will use fluorescence spectroscopy and other ion sensing methods to study if different carboxylates and phosphates can diffuse spontaneously across the membrane. Then, synthetic transporters will be added to the membranes, to study the rate at which these transport the different anions. You will analyse how the rate of diffusion and transport of the anions depends on their structure.

Prerequisite

Basic knowledge of organic chemistry and spectroscopy.

Contact person

For more information please contact : Hennie.Valkenier@ulb.be

references

Norvaisa 2024 Torres-Huerta 2025

Promotor, co-promotor, advisor : hennie.valkenier@ulb.be, - , Gianluca Weyckmans Mele

Research Unit : ENGINEERING OF MOLECULAR NANOSYSTEMS

Description

Project title

Study of cation ionophores with anticancer and antibacterial purposes.

Context

Copper(I) ions are important for living organisms, especially when incorporated in the active catalytic site of enzymes. Cu+ cannot diffuse through cell membranes spontaneously, but several membrane proteins take care of its transmembrane transport, a process that can be mimicked by small synthetic molecules. We have developed synthetic transporters for Cu+ and demonstrated these to be able to transport Cu+ into liposomes as model membranes, as monitored by the quenching of a fluorescent probe. The transport of Cu across cell membranes was then studied in yeast cell, where lead compound “Cuphoralix” restored the growth of cells that were modified by deleting their copper transport protein CTR1. Furthermore, Cuphoralix and closely related analogues were found to have very potent anti-cancer properties by our collaborators at the Université Grenoble Alpes.

Objective

Based on our recent result on the transport of copper cations, this project will focus on the activity and selectivity for different metal cations, such as zinc, iron, and silver. The aim of this project is to study different organic compounds as transporters for metal cations in liposomes as model systems. The results obtained will be then correlated to their potential anti-cancer activity.

Methods

This project will start with the preparation of liposomes with fluorescent probes encapsulated. The fluorescence of different probes will be studied, as well as their response upon the addition of different cations. Subsequently, the transport of the cations by the different organic compounds will be studied by fluorescence spectroscopy.

Prerequisite

General Chemistry Understanding of structural organic chemistry

Contact person

For more information please contact : Hennie.Valkenier@ulb.be

references

Copper transport across cell membranes by calix[4]arene-based cationophores leading to potent biological activity, ChemRxiv Preprint, 2025, https://doi.org/10.26434/chemrxiv-2025-2z3pd

Promotor, co-promotor, advisor : serge.schiffmann@ulb.be, Olivier Debeir, Christophe Varin

Research Unit : LAB. NEUROPHYSIOLOGY

Description

Title Analysis of large-scale calcium imaging recordings of neuronal activity in relation with naturalistic and acquired behaviors.

Context The basal ganglia play a key role in the control of both goal-directed behaviors and natural, self-paced behaviors. The proper initiation and execution of these behaviors rely heavily on appropriate functioning within the basal ganglia. Indeed, basal ganglia dysfunction is at the core of various disorders, including Parkinson’s disease, autism spectrum disorders, and schizophrenia (Gunaydin and Kreitzer, Annu. Rev. Physiol., 2016). The striatum, which is the main entry nucleus of the basal ganglia, consists of two types of striatal projection neurons (SPNs) that differ based on their expression of either dopamine D1 or D2 receptors and their respective direct or indirect projections to the output nuclei of the basal ganglia (dSPNs or iSPNs). Previous evidence has led to divergent conclusions on the respective engagement of both pathways during the execution of spontaneous actions as well as during the acquisition of new goal-directed instrumental behaviors. Our team recently proposed an updated model for motor encoding among SPNs in the dorsal striatum that relies on the congruent activation of dSPNs, which encode multiple accessible behaviors in a given context to promote these behaviors, and iSPNs, which encode for and inhibit competing behaviors (Varin et al., Nat. Comm., 2023). As a result, the coactivation of specific subsets of behavior-promoting dSPNs and behavior-suppressing iSPNs alongside specific inhibition of subsets of iSPNs allowing behavior expression would result in the selection and execution of only one motor program.

Project This work aims at validating and expanding our understanding of the organization of neuronal activity among dSPNs and iSPNs during naturalistic and learned goal-directed opreant behaviors using in vivo microendoscopic recordings of neuronal activity recorded through calcium indicators expressed specifically in dSPNs and iSPNs. The project will rely on the analysis of already acquired datasets obtained in mice learning an operant goal-directed task and on the acquisition and analysis of new recordings harvesting in mice submitted to a set of behavioural experiments (e.g. self-paced exploration of open fields of different shapes, sizes, proximal cues, elevated plus maze, light-dark room). The goal of the project will be to decipher and compare the encoding of actions between dSPNs and iSPNs and how their respective encoding properties for a given action evolve during learning and when external contingencies are modified.

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, jennifer.dhont@hubruxelles.be, younes.jourani@hubruxelles.be

Research Unit : LISA - IMAGE

Description

Project title

The project aims to solve an open issue in a certain domain of application.

Background

Tumor segmentation in digital pathology plays a crucial role in breast cancer diagnosis and prognosis [1], [2]. Precise delineation of malignant epithelial regions in hematoxylin and eosin (H&E)-stained or immunohistochemistry (IHC)-stained slides enables downstream analyses, such as cellularity estimation and biomarker quantification for diagnostic pathological examination, therapeutic response assessment, treatment selection, and survival prediction [3]–[8]. Deep learning-based segmentation approaches overcome the inefficiency of manual assessment, enabling high-throughput analysis of histopathological datasets. However, current approaches predominantly rely on supervised learning, which requires labor-intensive pixel-level manual annotations that are impractical at scale [9]–[11].

Weakly supervised learning has emerged as a promising alternative, leveraging coarse-grained labels to reduce annotation burdens. Yet, existing solutions are constrained by the restriction to whole-slide image (WSI)-level classification [12], reliance on partial cell-level annotations [13], and unproven generalizability across diverse breast cancer cohorts and staining protocols [14], [15]. These challenges underscore the need for a weakly supervised segmentation method that is trained using only image-level annotations while achieving pixel-level precision in malignant epithelium delineation and generalizing to heterogeneous breast cancer datasets.

Specific tasks

• Literature study to get familiar with the different topics.

• Perform data preprocessing, including extracting patches from whole slide images, applying color deconvolution to separate the Hematoxylin stain from H&E and IHC images using ImageJ, and applying data augmentation techniques such as flipping, rotation, and adjusting brightness and contrast to address class imbalance.

• Implement prevalent convolutional neural network (CNN) and Transformer models, as described in Table 4 and Table 5 of Ref. [16], and conduct training and inference of these models using Python, preferably with PyTorch.

• Validate the segmentation results predicted by these models across various breast cancer datasets, including H&E and IHC images, by comparing them to the ground truth segmentation mask (e.g., on the MHCI and BCSS datasets) or the ground truth cellularity (e.g., on the BreastPathQ and Post-NAT-BRCA datasets).

• [Optional] Develop multiple instance learning (MIL) techniques to improve segmentation performance across diverse breast cancer datasets, aiming to achieve accuracy comparable to that of supervised semantic segmentation methods.

Resources

• BreastPathQ dataset: a public dataset consisting of 69 H&E stained WSI collected from the resection specimens of 37 post-neoadjuvant therapy patients with invasive residual breast cancer. 2579 image patches with ROI of 512 × 512 pixels are manually annotated with estimated cellularity ranging between [0, 1].

• Other public datasets: https://github.com/maduc7/Histopathology-Datasets

• IHC datasets in NEOCHECKRAY. There are 109 IHC patches stained with an MHC-I antibody with pixel-level manual annotations.

Prerequisite

• Python

Contact persons

Dr. Ir. Jennifer Dhont (jennifer.dhont@hubruxelles.be), Head of Data Science & AI Research Unit at Hopital Universitaire de Bruxelles (Erasme campus)

Pr O. Debeir (olivier.debeir@ulb.be)

references

[1] D. Yan, X. Ju, et al., “Tumour stroma ratio is a potential predictor for 5-year disease-free survival in breast cancer,” BMC Cancer, vol. 22, no. 1, p. 1082, Oct. 2022.

[2] L. Priya C V, B. V G, V. B R, and S. Ramachandran, “Deep learning approaches for breast cancer detection in histopathology images: A review,” Cancer Biomarkers, vol. 40, no. 1, pp. 1–25, May 2024.

[3] M. Peikari, S. Salama, et al., “Automatic Cellularity Assessment from Post-Treated Breast Surgical Specimens,” Cytometry A, vol. 91, no. 11, pp. 1078–1087, Nov. 2017.

[4] S. Akbar, M. Peikari, et al., “Automated and Manual Quantification of Tumour Cellularity in Digital Slides for Tumour Burden Assessment,” Sci Rep, vol. 9, no. 1, p. 14099, Oct. 2019.

[5] X. Catteau, E. Zindy, et al., “Comparison Between Manual and Automated Assessment of Ki-67 in Breast Carcinoma: Test of a Simple Method in Daily Practice,” Technol Cancer Res Treat, vol. 22, p. 15330338231169603, Jan. 2023.

[6] E. H. Allott, S. M. Cohen, et al., “Performance of Three-Biomarker Immunohistochemistry for Intrinsic Breast Cancer Subtyping in the AMBER Consortium,” Cancer Epidemiology, Biomarkers & Prevention, vol. 25, no. 3, pp. 470–478, Mar. 2016.

[7] T. Vougiouklakis, B. J. Belovarac, et al., “The diagnostic utility of EZH2 H-score and Ki-67 index in non-invasive breast apocrine lesions,” Pathology - Research and Practice, vol. 216, no. 9, p. 153041, Sep. 2020.

attached pdf document

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, Julien.Cabay@ulb.be, Thomas.Vandamme@ulb.be

Research Unit : LISA-IMAGE

Description

RAG (Retrieval-Augmented Generation) for Patents

This project consists in the design, development, and testing of a RAG system (an AI chatbot with a specific knowledge base) for a dataset of patents.

Context

Patents are an invaluable economic asset, enabling inventors to protect their inventions for a set duration of time. Those invaluable assets, in the form of patent documents, represent an enormous challenge for the administrations responsible with the protection processes (i.e. Intellectual Property Offices). Those documents are highly technical, composed of different modalities (text and schematics), and are particularly numerous (there were more than 35 million patents in force worldwide as of 2023, source WIPO statistics database).

Recent technological advancements in the field of Artificial Intelligence (AI), namely Large Language Models (LLMs) and the chatbots that these power, carry enormous promises of automation for these complex tasks. One of those, Retrieval-Augmented Generation, is frequently branded as a solution to hallucination in LLMs, as well as enabling a relatively easy specialization of the model using a knowledge library.

Objective

In this project, you will design, develop and test such a solution on a large corpus of patents.

Methods

Different open-source LLMs can be used and benchmarked, as well as the different RAG techniques. The dataset can be sourced from Google Patents.

Prerequisite

• Python
• Machine Learning / Deep Learning

Contact person

For more information please contact : Thomas.Vandamme@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, Julien.Cabay@ulb.be, Thomas.Vandamme@ulb.be

Research Unit : LISA-IMAGE

Description

Design and Implementation of a viewer for IP (Intellectual Property) datasets

This project consists in the design, development, and implementation of a viewer website/software for IP datasets (Trade Marks, Patents, ...). The viewer will enable users, developers and researchers to search, label and extract different relevant aspects of the datasets.

Context

Current dataset viewer tools, such as label studio (https://labelstud.io/), have demonstrated their relevance in research and development ecosystems, especially those related to data and deep learning. However, those tools are not perfect, and several panes of datasets, such as those related to the legal field (esp. text documents) are left out of such solutions.

Objective

In this project, you will develop a complete tool (ideally web-based), or an open-source plug-in for another viewer/labelizer (such as label studio, for example), capable of handling multimodal informations, such as those relative to IP (e.g. images, 3D volumes, schematics, text, sound, ...).

Prerequisite

• Web Technologies
• Python

Contact person

For more information please contact : Thomas.Vandamme@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, Julien.Cabay@ulb.be, Thomas.Vandamme@ulb.be

Research Unit : LISA-IMAGE

Description

Automated web scraping for dataset compilations

This project consists in the design, development, implementation and testing of a series of automated web scrapers. The ultimate goal is to develop a series of tools to enable the acquisition and synchronisation of different datasets.

Context

Deep Learning relies on voluminous and (ideally) good-quality datasets. Those are, unfortunately, hard to gather and label.

In the field of Intellectual Property (IP, including, e.g. Patents, Trade Marks, Designs), some relevant informations are curated by public IP offices (tasked with the administration of the different associated rights; registration, protection, ...). Those public bodies make publicly available a series of information through various search engines (for example, see https://www.euipo.europa.eu/en/search and https://ipportal.wipo.int/home). The offices do not allow for bulk download, and curating by hand these datasets is a particularly tedious task (there are millions of registered rights, for example).

Objective

In this project, you will develop tools to enable the fast development of web scrapers, implementing various measures to disable anti-scraping protections on the websites. A new, untested use case will be chosen to illustrate the tools capabilities.

Methods

You will create web-based applications, interface with (simple) databases and external providers if needed. Your end-product will enable non-developers to choose the elements they want to retrieve automatically in a webpage, and various other settings.

Prerequisite

• Web Technologies
• Python
• (optional) Selenium or other automation software

Contact person

For more information please contact : Thomas.Vandamme@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, Julien.Cabay@ulb.be, Thomas.Vandamme@ulb.be

Research Unit : LISA-IMAGE

Description

Assessing legally-relevant similarities between goods and services of Trade Marks using Large Language Models (LLMs)

In this project, you will evaluate the capabilities of current state-of-the-art AI models to assess similarity between goods and services in the field of Trade Marks.

Context

In the field of Trade Mark (TM) Law, similarity is paramount. TMs that are too similar (under a legal criterion called the "Likelihood of Confusion", LoC) to another already registered TM may not be registered.

The LoC test, a complex and multifactorial assessment performed by judicial authorities, involves two sub-assessments: the similarities between the signs (the images or text of the TM, e.g. "Coca-Cola", or the "ULB" logo), and the similarities between the goods and services for which those are registered (e.g. "cold beverages", and "Higher Education").

Technological solutions have been developed to alleviate the administrative burden caused by the massive amount of registered TMs combined with ever-increasing TM applications. In the case of estimating the similarities between 2D signs, the arguably most promising solution is search engines, while in the case of assessing such similarities between goods and services (an equally important test), this solution could take the form of a classifier or of a semantic search system.

Objective

For this project, you will investigate the extent to which current AI LLMs are able to assess this legal test of similarities of goods and services, conclude on the state of the technology for these tasks, and propose avenues to increase the performances.

Methods

A dataset of decisions, of which you will be able to extract the portion relevant to the assessment of similarity between goods and services, is already available.

A possible course of action for the experimentation involves a first approach of zero-shot learning, possibly combined with Retrieval-Augmented Generation (RAG) and/or mixture of experts. Strong of your early experiments, you will investigate hands-on the capacity of those tools by designing and training your own AI models to solve this task specifically, potentially based on pre-trained LLMs.

Prerequisite

• Machine Learning / Deep Learning
• Python

Contact person

For more information please contact : Thomas.Vandamme@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, Julien.Cabay@ulb.be, Thomas.Vandamme@ulb.be

Research Unit : LISA-IMAGE

Description

AI Perception of Design Similarity: 2D Views versus 3D Designs

In this project, you will investigate technico-legal considerations relative to AI algorithms and how they perceive objects. To do this, you will design, implement, train and test various algorithms and AI models to perform a variety of atomic tasks.

Context

Designs are an important part of Intellectual Property, protecting the appearance of goods (e.g. the shape of the Coca-Cola bottle). These designs must be registered to be granted protection, and their registration is done through a series of 2D views (of these 3D objects), for administrative reasons.

In an infringement context, a judicial authority must assess whether two designs differ sufficiently or not. This authority is capable of inferring, from several 2D views, the general shape of the object, in order to perform this assessment.

Recent advances in AI have enabled various tools in this field, most importantly a design search tool, that enables individuals seeking registration of their design to search for potentially infringing designs. Those private tools are probably based on these 2D views, and may therefore lack a 3D understanding.

Objective

In this master thesis, you will investigate whether Design search tools indeed suffer from this 2D/3D semantic gap, and if they are able to accurately implement the legal rules pertaining to this field. Additionnally, you will inform on whether judicial authorities do indeed base themselves on the 3D view, or if the 2D views are sufficient for those tasks.

Methods

You will develop different AI models capable of inferring 2D views from a 3D mesh object (1), of creating a 3D mesh from different 2D views (2), and finally, implement a research protocol to test the hypotheses.

Prerequisite

• Machine Learning / Deep Learning
• Python

Contact person

For more information please contact : Thomas.Vandamme@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, adrien.foucart@ulb.be, arthur.elskens@ulb.be

Research Unit : LISA-IMAGE

Description

Evaluation of a Feature-Based Registration Pipeline for Whole-Slide Images

This project focuses on the evaluation of various configurations of a feature-based registration pipeline, with particular attention to their robustness under varying initial conditions in the domain of Whole-Slide Image (WSI) registration.

Context

In digital pathology, the integration of information from multiple WSIs is often required to obtain a comprehensive understanding of complex biological processes, such as cancer development and progression. This integration requires accurate spatial alignment of corresponding tissue regions across WSIs, a process referred to as WSI registration. Due to challenges such as image artefacts, varying staining, and large tissue deformations, WSI registration remains a difficult problem in computer vision.

Recent challenges, such as ANHIR [Borovec2020] and ACROBAT [Weitz2024], have brought attention to the state-of-the-art in WSI registration. These challenges evaluate the performance of registration algorithms on real-world datasets comprising multi-stained slides, such as Hematoxylin & Eosin and Immunohistochemistry. Successful approaches generally follow a two-stage process: an initial low-resolution rigid or affine alignment, followed by a high-resolution non-rigid (deformable) registration. This two-step approach effectively constrains the parameter space within which the non-linear elastic transformation is subsequently estimated, thereby significantly improving the quality of the final alignment.

Among the strategies for estimating initial alignment, feature-based methods have emerged as the most widely used, with six out of eight top-performing teams in the ACROBAT challenge employing such techniques [Weitz2024]. Feature-based registration typically involves three key stages: (i) pre-processing, (ii) local feature extraction, and (iii) robust matching. Each of these stages can be performed using a variety of algorithms, ranging from traditional approaches to deep learning-based methods [Marzahl2021, Gatenbee2023, Elskens2023, Wodzinski2024, Elskens2025].

Objectives

The primary objective of this project is to:

• Evaluate different configurations (robust matching stage) of feature-based registration pipeline.
• Assess the robustness of different configurations of feature-based registration pipelines to diverse initial conditions, including large spatial displacements and staining artefacts.

Methods

The project will involve a comparative study of several state-of-the-art feature detection and matching algorithms found in recent literature. These may include: SuperPoint [DeTone2018], LightGlue [Lindenberger2023], LoFTR [Sun2021] and OmniGlue [Jiang2024]. Each algorithm will be implemented and evaluated under controlled experimental settings.

A novel evaluation protocol will be developed, grounded in existing metrics from the literature while integrating innovative criteria specifically tailored to the tasks explored in this project. The evaluation will include, but not be limited to, assessments of robustness to large displacements (e.g., rotation, translation, and shear) as well as the ability of each pipeline configuration to effectively filter outliers during robust matching. The goal is to design a comprehensive and reproducible benchmarking approach focused on critical aspects of feature-based registration pipelines within the context of digital pathology.

Prerequisites

Candidates should have:

• A solid foundation in Python programming
• A willingness to work with containerization tools such as Docker (prior experience is a plus but not mandatory)
• Successfully completed INFO-H500 or an equivalent course in image processing, computer vision, or machine learning

Contact person

For further information or to express interest in this project, please contact: arthur.elskens@ulb.be, adrien.foucart@ulb.be and olivier.debeir@ulb.be.

References

[Borovec2020] Borovec, J., Kybic, J., Arganda-Carreras, I., Sorokin, D.V., Bueno, G., Khvostikov, A.V., . . . Munoz-Barrutia, A. (2020, October). ANHIR: Automatic Non-Rigid Histological Image Registration Challenge IEEE Transactions on Medical Imaging, 39 (10), 3042-3052, https://doi.org/10.1109/TMI.2020.2986331

[Weitz2024] Weitz, P., Valkonen, M., Solorzano, L., Carr, C., Kartasalo, K., Boissin, C., . . . Rantalainen, M. (2024, October). The ACROBAT 2022 challenge: Automatic registration of breast cancer tissue. Medical Image Analysis, 97, 103257, https://doi.org/10.1016/j.media.2024.103257

[Marzahl2021] Marzahl, C., Wilm, F., Dressler, F., Tharun, L., Perner, S., Bertram, C., . . . Breininger, K. (2021). Robust Quad-Tree based Registration on Whole Slide Images. Proceedings of Machine Learning Research (Vol. 156, pp. 181-190).

[Gatenbee2023] Gatenbee, C.D., Baker, A.-M., Prabhakaran, S., Swinyard, O., Slebos, R.J.C., Mandal, G., . . . Anderson, A.R.A. (2023, July). Virtual alignment of pathology image series for multi-gigapixel whole slide images. Nature Communications, 14 (1), 4502, https://doi.org/10.1038/s41467-023-40218-9

[Elskens2023] Elskens, A., Foucart, A., Zindy, E., Debeir, O., Decaestecker, C. (2023, November). Assessing Local Descriptors for Feature-Based Registration of Whole-Slide Images. 2023 19th International Symposium on Medical Information Processing and Analysis (SIPAIM) (pp. 1-4). Mexico City, Mexico: IEEE. https://doi.org/10.1109/SIPAIM56729.2023.10373514

[Wodzinski2024] Wodzinski, M., Marini, N., Atzori, M., Müller, H. (2024, June). RegWSI: Whole slide image registration using combined deep feature- and intensity-based methods: Winner of the ACROBAT 2023 challenge. Computer Methods and Programs in Biomedicine, 250, 108187, https://doi.org/10.1016/j.cmpb.2024.108187

[Elskens2025] Elskens, A., Foucart, A., Debeir, O., Decaestecker, C. (2025, May). Impact of Pre processing and Local Feature Extraction on Feature Based Registration of Whole Slide Images. Preprint at http://dx.doi.org/10.13140/RG.2.2.29399.79521

[DeTone2018] DeTone, D., Malisiewicz, T., Rabinovich, A. (2018, June). SuperPoint: Self-Supervised Interest Point Detection and Description. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) (pp. 337-33712). Salt Lake City, UT, USA: IEEE. https://doi.org/10.1109/CVPRW.2018.00060

[Lindenberger2023] Lindenberger, P., Sarlin, P.-E., Pollefeys, M. (2023, October). LightGlue: Local Feature Matching at Light Speed. 2023 IEEE/CVF International Conference on Computer Vision (ICCV) (pp. 17581-17592). https://doi.org/10.1109/ICCV51070.2023.01616

[Sun2021] Sun, J., Shen, Z., Wang, Y., Bao, H., Zhou, X. (2021, June). LoFTR: Detector-Free Local Feature Matching with Transformers. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 8918-8927). Nashville, TN, USA: IEEE. https://doi.org/10.1109/CVPR46437.2021.00881

[Jiang2024] Jiang, H., Karpur, A., Cao, B., Huang, Q., Araujo, A. (2024, June). OmniGlue: Generalizable Feature Matching with Foundation Model Guidance. 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) (pp. 19865-19875). Seattle, WA, USA: IEEE. https://doi.org/10.1109/CVPR52733.2024.01878

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, feras.almasri@ulb.be,

Research Unit : LISA-IMAGE

Description

Automated Interpretation of P&ID Drawings

The project aims to address a key challenge in industrial automation by developing AI and computer vision methods to automatically interpret and digitize Piping and Instrumentation Diagrams (P&IDs).

Context

The project is carried out in collaboration with Engie and TRACTEBEL.

Annotated P&ID drawings are already available for model training and validation.

Objective

The objective is to develop a robust pipeline capable of detecting and recognizing symbols and text, associating them based on their spatial and contextual relationships, and detecting and tracking lines within P&ID drawings. The ultimate goal is to produce a structured digital representation of the diagrams for use in automation and further analysis.

Methods

Various methods will be explored, including convolutional neural networks (CNNs) for symbol and text recognition, and specialized algorithms for line and path detection. Post-processing techniques will be designed to accurately associate and structure the extracted elements.

Prerequisite

• Image processing

• Deep Learning

• Python

Contact person

For more information please contact : feras.alamsri@ulb.be, olivier.debeir@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, feras.almasri@ulb.be,

Research Unit : LISA-IMAGE

Description

User Behavior Analysis via Tweet Detection and Interaction Tracking in Mobile Screen Recordings

The project aims to explore user interaction patterns on social media by detecting tweets in screen recordings, recognizing tweet content, and tracking user actions such as clicks and follows.

Context

The project is conducted in collaboration with the faculty of psychology.

A dataset of mobile screen recordings is available, capturing real user interactions on platforms like Twitter.

### Objective

The goal is to develop a system that can detect tweets within mobile screen recordings, apply text recognition to extract tweet content, and track user interactions (e.g., clicks, scrolling, and follows). This information will help study behavior patterns and engagement strategies on social media.

### Methods

Multiple techniques will be tested, including object detection for tweet localization, OCR methods for text recognition, and temporal tracking algorithms to monitor user interactions over time. Event detection and classification will also be explored to analyze engagement behavior.

Prerequisite

• Image processing

• Deep Learning

• Python

• Flask/FAST API

• Docker

Contact person

For more information please contact: feras.almasri@ulb.be, olivier.debeir@ulb.be

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, erwin.woff@ulb.be,

Research Unit : LISA-IMAGE

Description

Multimodal Deep Learning for Deauville Score Prediction in Lymphoma Using [18F]FDG-PET/CT Imaging and Clinical Reports

Subject:

Advancing Automated Deauville Scoring Through Vision-Language Models

Description:

Context and Aim of the Project:

The Deauville Score (DS) is a critical tool in the assessment of treatment response in Hodgkin and non-Hodgkin lymphoma, based on [18F]FDG-PET/CT imaging. In current clinical practice, DS assignment is done visually by physicians, which is time-consuming and subject to interobserver variability. While deep learning models trained on PET maximum intensity projections (MIPs) have shown strong potential for binary classification of DS (1–3 vs. 4–5), challenges remain in achieving higher granularity, robustness, and clinical interpretability.

Recent advances in large language models and domain-adapted transformers (e.g., BioClinicalBERT, RadBERT) have shown promise in extracting Deauville scores directly from nuclear medicine reports (Hueman et al., 2023). Building on this, the goal of this project is to explore multimodal deep learning architectures that integrate PET images and their corresponding medical reports to enhance Deauville score classification. However, using image-only or text-only architectures remains a viable and simpler baseline, the student will have the flexibility to pursue a purely visual or a multimodal path depending on interest and feasibility.

Ultimately, the core objective is to improve the robustness and granularity of Deauville score predictions, potentially moving from binary to full five-class classification. The thesis also opens opportunities to explore interpretability techniques, such as lesion localization or clinical explanation generation, and sets the stage for external validation and eventual clinical deployment.

Objectives:

1. Improve Deauville score prediction:

   • Move beyond binary classification toward full five-class DS prediction by combining PET imaging and text reports.

2. Integrate vision and language modalities:

   • Extend the current framework with transformer-based encoders (e.g., BioClinicalBERT) for the report and CNN or ViT-based encoders for the image, evaluating different fusion strategies. Pretrained models from Huemann et al. are available as a starting point.

3. Enhance interpretability and clinical insight (bonus)

   • Add auxiliary tasks such as lesion localization or lesion description generation to improve model explainability and potentially accuracy.

4. Evaluate Generalization and Clinical Usefulness:

   • Possibility to test the model performance on potential external cohorts for validation.
   • Compare predictions with expert-assigned scores and evaluate agreement.

Methodology:

• Use existing anonymized FDG-PET/CT data and matched reports from the Institut Jules Bordet.
• Preprocess MIP images and radiology reports using standardized pipelines.
• Fine-tune pre-trained BERT-based models on nuclear medicine language, guided by work such as Huemann et al. (2023) or directly use their domain adapted model available on Hugging face.
• Experiment with text and image-text models using MIP images and transformer-encoded reports.
• Evaluate performance on internal and (if available) external validation cohorts.

Available Resources:

• Annotated dataset with Deauville scores and clinical reports (in French).
• Existing codebase and models from the prior thesis with preprocessing scripts and baseline models for PET MIPs (available on GitHub: Alichnikof/Deauville_DeepLearning).
• Huemann et al.’s code for report processing and fine-tuning (zhuemann/NuclearMedicineDomain_Adaptation).
• Preprocessed anonymized dataset with corresponding clinical reports.
• Scripts for visualization, evaluation, and data augmentation.
• Potential future datasets enabling external validation.

Expected Outcomes:

• A refined and possibly multimodal AI model for Deauville score classification.
• Deeper understanding of the complementarity of PET imaging and narrative reports.
• Foundations for future applications in AI-assisted lymphoma response assessment and model interpretability.

Supervision and Collaboration:

This project will be conducted under the supervision of Prof. Olivier Debeir (ULB) and Prof. Erwin Woff (Institut Jules Bordet), fostering interdisciplinary collaboration between biomedical engineering and nuclear medicine.

Key References

• Mezher A. Deep Learning for Binary Deauville Scoring in Lymphoma [18F]FDG-PET/CT: Transfer Learning and External Validation, ULB 2025.
• Huemann Z. et al., Domain-adapted large language models for classifying nuclear medicine reports, NPJ Digit. Med. 2023.
• Häggström I. et al., Deep learning for [18F]FDG-PET/CT classification in patients with lymphoma, Lancet Digit. Health, 2023.

Promotor, co-promotor, advisor : olivier.debeir@ulb.be, - ,

Research Unit : LISA-IMAGE

Description

Transformers and Attention Maps for Semantic Analysis of Vectorized Engineering Blueprints

Context

This project addresses an ongoing challenge in the field of automatic document analysis. While Optical Character Recognition (OCR) is now widely regarded as a solved problem, the automated interpretation of more complex documents—particularly technical and engineering drawings—remains difficult.

Recent advances in machine learning have enabled higher-level abstraction in the analysis of image-based documents. Most existing methods for blueprint interpretation adopt raster-based approaches, analyzing 2D pixel representations of documents. However, many technical drawings are also available in vector format (e.g., as embedded objects in PDF files). Although these vector representations are typically designed for visualization and printing, and thus contain limited semantic metadata, they may offer structural advantages that can be exploited for semantic analysis.

Objective

The primary objective of this project is to evaluate the potential advantages of vector-based analysis over traditional raster-based approaches in the semantic interpretation of engineering blueprints.

Methods

A recent study by Carrara et al. [Carrara2024] proposes a novel vector-based framework leveraging Convolutional Neural Networks (CNNs) and Graph Attention Networks (GATs). This project will build upon that work to further investigate the role of attention mechanisms and transformer architectures in processing vectorized technical documents.

A publicly available annotated dataset will serve as the foundation for experiments. Additional real-world examples provided through industrial collaboration will be used for further testing and validation.

Prerequisites

• Proficiency in Python programming
• Knowledge of deep neural networks (DNNs) and attention-based models

Contact

For further information, please contact: olivier.debeir@ulb.be

References

Carrara, Andrea, Stavros Nousias, and André Borrmann. “VectorGraphNET: Graph Attention Networks for Accurate Segmentation of Complex Technical Drawings.” arXiv preprint arXiv:2410.01336 (2024). https://arxiv.org/pdf/2410.01336

Promotor, co-promotor, advisor : bram.vanderborght@vub.be, Ellen Roels, Ellen Roels

Research Unit : MECH

Description

Designing programmable soft matter with tunable bending behaviour

Context

Soft robotics is a subfield of robotics that leverages the compliance of soft materials such as silicone rubber to create flexible and adaptive robotic systems. Unlike traitional rigid robots, soft robots can deform continuously, enabling new types of locomotion, manipulation and interaction with unstructured environments. One aproach for controlling deformation in soft robots is tendon-driven actuation, where embedded tendons are pulled to transfer forces. By carefully designing the geometry of these structures, it is possible to tune the deformation behaviour.

Objective

This thesis focusses on the development of soft units with tunable bending behaviour, actuated using a tendon. By modifying their geometrical parameters, the bendng behaviour can be tuned. These units, designed as cubic voxels are stacked on top of each other, and actuated using a single tendon routed through them. When the tendon is pulled, the units should bend sequentially in a controlled manner. Upon releasing the tendon, the unbending should follow a different, predefined sequence, introducing an asymmetric deformation-recovery cycle.

Methods

As part of this project, you will: - Design and optimize the soft units, exploring how geometrical features influence their bending behaviour. - Develop a computational model using finite element analysis (preferably Abaqus) to simulate their mechanical response and predict the deformation sequence. - Manufacture the units by casting them in silicone rubber. - Experimentally validate their behaviour, analyzing the deformation pattern and actuation force to compare with simulation results. - Explore if this concept can be used as legs for a walking exploration robot.

Prerequisite

There are no prerequisites, but prior Abaqus experience can be useful. Estimated workload: Literature study: 25% analysis (FEM, calculations): 30% design: 20% manufacturing: 10% experiments: 15%

Contact person

For more information please contact : ellen.roels@vub.be

attached pdf document

Promotor, co-promotor, advisor : bram.vanderborght@vub.be, Ellen Roels, Ellen Roels

Research Unit : MECH

Description

Directional friction surfaces inspired by biological microstructures

Context

In nature, many organisms have surface structures that enable direction-dependent control over friction. For example, snakes use overlapping scales to generate high friction in one direction while enabling smooth forward motion in another. Similarly, sharks use micoscopic indents to reduce drag and enhance swimming efficiency. Studying and mimicking these biological strategies can inspire new materials with (tunable) directional friction, which can be useful for applications in robotic grippers, wearable robotics, or locomotion.

Objective

The goal of this thesis is to design and fabricate soft surfaces that exhibit direction-dependent friction properties. You will study how the geometric parameters influence friction and fabricate your designs using 3D printing or silicone moulding.

Methods

For this thesis, you will first start with a literature study. Based on this, you will create a design for a (tunable) directional friction surface, for which you will do a parametric study in simulation to see the influence of various geometrical parameters. With this knowledge, you will fabricate several of your designs and verify their properties experimentally.

Prerequisite

There are no prerequisites, but prior finite element simulation experience can be useful.

Contact person

For more information please contact : ellen.roels@vub.be

attached pdf document

Promotor, co-promotor, advisor : nicolas.pauly@ulb.be, - , Dhont Jennifer

Research Unit : MÉTROLOGIE NUCLÉAIRE

Description

Modelling lymph node involvement in early stage breast cancer

Background The treatment landscape for patients with early-stage breast cancer has evolved significantly over the last few decades, aiming for optimal outcomes while minimizing treatment-related harm [1]. Advances in targeted systemic therapies and technological innovations in surgery and radiotherapy have markedly improved patient outcomes [2]. Specifically, in several meta-analyses, radiotherapy to the locoregional lymph node levels after surgery has demonstrated a reduction in disease recurrences and breast cancer mortality, especially in early-stage breast cancer patients with positive lymph nodes [3].

Despite these benefits, patients may encounter side effects during radiotherapy, such as skin irritation, fatigue, and breast edema. Potential long-term complications include changes to the breast, arm and shoulder pain, and moderate to severe fibrosis, which can affect the quality of life substantially [4,5]. Less frequently, but potentially more serious long-term complications include heart and lung diseases and the development of secondary cancers [6]. The probability and severity of these side effects are often directly correlated to the volume of irradiated tissue, emphasizing the delicate balance in defining appropriate radiotherapy target volumes. The objective of this research project is to optimize the target volumes for locoregional radiotherapy in patients with early-stage breast cancer, through big data analysis and modelling of lymph node involvement using Hidden Markov Models.

Specific Tasks • Perform a literature study to get familiar with the different topics. • Perform descriptive and statistical data analysis on a large (N>500), retrospective cohort of early-stage breast cancer patients, similar to [7]. The data will include patient and disease characteristics (clinical data), treatment-related data and lymph node involvement obtained from different modalities (imaging and pathology). Data analysis will be carried out in python using pandas library, and libraries for data visualization. • Get familiar with an already implemented Hidden Markov Model (in Python), including the dataset used for fitting. • Adapt the Hidden Markov Model to include patient- or disease-specific parameters, such as primary tumour location. • Evaluation of new model performance after adaptation and refitting the model.

Contact • Dr. Ir. Jennifer Dhont (jennifer.dhont@hubruxelles.be), Head of Data Science & AI Research Unit at Hopital Universitaire de Bruxelles (Erasme campus)

References 1. Waks, A. G., et al. Breast Cancer Treatment: A Review. JAMA 321, 288–300 (2019). 2. Bray, F., et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 68, 394–424 (2018). 3. Abe, O., et al. Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival. Lancet 366, 2087–2106 (2005). 4. Hopwood, P., et al. Comparison of patient-reported breast, arm, and shoulder symptoms and body image after radiotherapy for early breast cancer. Lancet Oncol 11, 231–240 (2010). 5. Lyngholm, C. D., et al. Long-term follow-up of late morbidity, cosmetic outcome and body image after breast conserving therapy. Acta Oncol (Madr) 52, 259–269 (2013). 6. Piroth, M. D., et al. Heart toxicity from breast cancer radiotherapy: Current findings, assessment, and prevention. Strahlentherapie Und Onkologie 195, 1 (2019). 7. Ludwig, R., et al. Detailed patient-individual reporting of lymph node involvement in oropharyngeal squamous cell carcinoma with an online interface. Radiother Oncol 169, 1-7 (2022).

attached pdf document

Promotor, co-promotor, advisor : francois.horlin@ulb.be, Jean-Francois Determe,

Research Unit : OPERA - WIRELESS

Description

Context:

Wi-Fi modems are continuously evolving to meet the ever-increasing expectations of the users in terms of communications rates. The last amendment of the 802.11 standard, the 11be amendment referred to as extremely high throughput (EHT), specifies the Wi-Fi communications at frequencies below 7 GHz on a 320 MHz-wide bandwidth and by using up to 16 spatial streams created by arrays of antennas. Communication rates up to 50 Gbps will be supported by your modem!

At the same time, the new amendment 11bf is also developed to support Wi-Fi sensing besides communications. The principle is to leverage the channel measurements already useful for the communications and characterizing the environment to offer new breakthrough applications aiming at improving the autonomy and the security at home. Wi-Fi sensing can be seen as a follow-up of the effort on the design of Wi-Fi based passive radars, i.e. radars working by observing the Wi-Fi communications signals of opportunity to sense the environment.

Objective:

Among the envisioned Wi-Fi sensing use cases, the monitoring of activities to assess the health state or the fast detection of specific events like a fall to quickly call emergency services may help developing the autonomous living of the elderly at home. This calls for the estimation of parameters like the walking speed and the deployment of classification algorithms to detect specific events. The goal of this master thesis is to design a Wi-Fi based passive radar to track people indoors and extract micro-Doppler signatures useful to detect events like a fall or other activities when inputted to a classification algorithm.

Methodology and tools: • Design of Wi-Fi based radar system People tracking indoors based on range/Doppler/angle estimations Micro-Doppler signature extraction based on time-frequency analysis • Settle hardware setup and acquire experimental data (USRP X310/X410) • Conceive a Wi-Fi based health monitoring system Physical parameter estimation to quantify level of activity (actimetry) Classification algorithms for fall detection incl. feature extraction

Contact: francois.horlin@ulb.be

attached pdf document

Promotor, co-promotor, advisor : bram.vanderborght@vub.be, - , Pasquale Ferrentino

Research Unit : R&MM

Description

Project title

The project aims to solve an open issue in a certain domain of application.

Context

The project is done in collaboration with This Company.

Many data are already available

Objective

Here the goal

Methods

Different methods are to be tested...[Ronneberger2015]

Prerequisite

• C,
• C++,
• Python

Contact person

For more information please contact : contact@ulb.be

references

Ronneberger2015

attached pdf document

Promotor, co-promotor, advisor : ilias.el.makrini@vub.be, - , Tom Turcksin

Research Unit : R&MM

Description

Advanced Calibration Methods for EMG: Developing a Customizable Calibration Framework

The project aims to improve the accuracy and relevance of EMG-based assessments by developing a flexible, task-specific calibration framework tailored to human movement analysis.

Context

The project is done in collaboration with AugmentX (VUB). It leverages Cometa and TMSI EMG sensors available at the AugmentX infrastructure (www.augmentx.be). Many data are already available to support development and validation.

Objective

To develop a customizable EMG calibration and normalization framework that offers experimenters multiple task-relevant calibration methods. The focus will be on the elbow joint, incorporating muscle fatigue and joint angle considerations, with the ultimate goal of estimating joint torque from EMG data.

Methods

Different methods are to be tested and developed, namely:

• Literature review of existing EMG calibration techniques;

• Selection and implementation of task-relevant methods;

• Integration of muscle fatigue and joint kinematics into calibration;

• Development of a GUI or script to guide users through calibration steps;

• Testing and validation using Cometa and TMSI sensors.

Prerequisite

• Matlab,
• Python or C++

Contact person

For more information please contact : tom.turcksin@vub.be

Promotor, co-promotor, advisor : tom.verstraten@vub.be, Prof. Dr. ir. Bram Vanderborght (VUB), Amin Khorasani

Research Unit : R&MM

Description

Promotor Prof. Dr. ir. Tom Verstraten (VUB)

Co-Promotor Prof. Dr. ir. Bram Vanderborght (VUB)

Advisor Amin Khorasani

Laboratory/research unit Brubotics, R&MM

Subject title Patient-Specific 3D Eye Modeling and Tool Path Simulation for Ophthalmic Vitreoretinal Surgery Planning:

In ophthalmic surgeries such as vitrectomy or intravitreal injections, tool maneuvering inside the eye must be extremely precise to avoid damage to delicate structures like the retina. While robotic systems offer tremor elimination and sub-millimeter accuracy, surgeons lack clear feedback about where the tool tip is inside the eye. This problem becomes more complex as visualization during these procedures is limited to a narrow microscopic 2D field. This thesis proposes the development of a simulation-based solution that provides surgeons and researchers with a visual estimation of the tool path inside the eye. By combining pre-operative data such as OCT scans (cross section images of retina) and IOL Master biometry (axial length, white-to-white, anterior chamber depth), a patient-specific 3D model of the eye can be generated. Once the trocar is inserted and its position is tracked on the eye, this entry point becomes the pivot for simulating the internal trajectory of the surgical tool. This enables estimation of the tool tip location in real time and supports safer, more informed operations.

Objectives: 3D Eye Model Generation: Develop an algorithm to generate a geometrically realistic 3D eye model from patient-specific inputs (image processing of OCT slices, IOL Master data).

Tool Path Simulation: Implement a method to simulate the internal path of a surgical instrument inserted through a trocar, assuming pivoting around the scleral entry point. Visualization & Surgeon Feedback: Build a user-friendly visualization layer that maps tool movement onto the virtual eye model. Validate clarity and usability with input from ophthalmic surgeons.

Research Phases: Phase 1 (25%): Background study on eye anatomy, biometric measurements, and surgical workflow. Phase 2 (45%): Development of the 3D eye model (Integration of biometric measurements and oct images) and simulation of tool motion around the trocar. Phase 3 (30%): Visualization interface design and scenario simulation based on realistic clinical inputs. Learning Outcomes: The student will gain knowledge in image processing, 3D modeling, anatomical interpretation, and constrained tool kinematics. They will also build skills in Python programming, visualization libraries (e.g., VTK or PyVista), and simulation of minimally invasive tool motion. This project provides foundational insights into pre-operative planning tools for ophthalmic surgery and helps bridge the gap between clinical imaging and digital simulation. This project will be conducted in collaboration with UZ Brussel hospital. Interest and Contact: Are you interested in contributing to this cutting-edge research in robotic assisted surgery, image processing, and 3D modeling? Contact me at Amin.khorasani@vub.be for more information.

attached pdf document

Promotor, co-promotor, advisor : tom.verstraten@vub.be, Prof. Dr. ir. Bram Vanderborght, Amin Khorasani

Research Unit : R&MM

Description

Transparent, Backdrivable, and Reliable Actuation for Robot-Assisted Surgery: In robot-assisted and haptic-enabled surgical systems, the actuator is central to achieving high transparency, accurate force rendering, and safe tool-tissue interaction. Transparency, the ability to convey tactile cues to the operator with minimal distortion, is critical in microsurgical procedures where forces are often close or even below human perception thresholds, and minor deviations can have serious consequences. This demands actuators with low friction, minimal impedance, high control bandwidth, and mechanical precision. Simultaneously, these systems must be reliable over repeated use, resistant to degradation, and suitable for compact integration into clinical tools.

A variety of actuation mechanisms have been studied and implemented in surgical and haptic devices. Cable and tendon driven systems are widely used to enable remote actuation, reducing distal mass and improving maneuverability, particularly in handheld or minimally invasive tools. Direct-drive motors offer excellent backdrivability and precise control, making them well-suited for micromanipulators, although they are often limited in terms of torque density and size. Piezoelectric and ultrasonic actuators, known for their compactness and high-frequency response, have been deployed in ophthalmic and neurosurgical instruments for applications requiring sub-micron positioning. Fluidic actuators, including pneumatic and hydraulic systems, provide high force outputs in compact configurations, although they pose challenges in terms of controllability and dynamic response.

This proposal also explores hybrid actuation systems that combine stiff, high-resolution position controlled actuators with torque/force controlled mechanisms. These configurations aim to exploit the benefits of both domains offering ultra-precise positioning for steady manipulation and compliant interaction for force-sensitive environments. Such systems are particularly promising for scenarios requiring both stability near static targets and responsiveness to dynamic interactions, such as intraocular injections or membrane peeling. This thesis will analyze and compare these actuation strategies through modeling, simulation, and experimental validation, focusing on their transparency, reliability, and suitability for clinical integration in robotic-assisted surgical platforms.

Objectives: Survey: Review of actuation mechanisms suitable for haptic/robot-assisted surgery. Modeling: Simulate actuator dynamics and transparency using MATLAB/Simulink or equivalent tools. Reliability Study: Analyze degradation effects (e.g., backlash, friction) and their influence on force rendering. Experimental Validation: Build and test a prototype actuator with high backdrivability and transparency for an application like needle insertion. Design Recommendations: Summarize insights into how actuator designs can be optimized for surgery.

Research Phases: Phase 1 (20%): Literature review and definition of metrics for transparency, backdrivability, and reliability. Selection of one or more actuator types for study.

Phase 2 (40%): Simulation and analysis of selected actuator types. Study dynamic behavior, control bandwidth, and response under failure-like conditions (e.g., added friction or backlash). Phase 3 (40%): Prototyping and experimental validation of a selected actuator. Measurement of transparency, impedance, and force feedback fidelity using sensors. Evaluate robustness under repeated loading and identify improvement strategies.

Learning Outcomes: The student will gain a thorough understanding of actuator behavior in high-precision robotic systems and their role in safety-critical applications like surgery. They will develop skills in: Modeling and simulating mechatronic systems. Understanding haptic system requirements and evaluation metrics. Designing and controlling backdrivable actuators. Performing hands-on experimental validation, including sensor integration, data acquisition, and performance testing. Interpreting experimental results to guide design improvements.

Interest and Contact: Are you interested in contributing to the next generation of surgical robotic systems by exploring the heart of haptic feedback, and precision actuators? Contact me at Amin.khorasani@vub.be for more information.

attached pdf document

Promotor, co-promotor, advisor : tom.verstraten@vub.be, - , M Wu

Research Unit : ROBOTICS AND MULTIBODY MECHANICS RESEARCH GROUP

Description

Context

One of the most critical challenges for development of exoskeletons and exosuits is the design of physical attachments – mechanical straps, cuffs, or sleeves – that connect the robot to the human user. In addition to creating a mechanical connection between the exoskeleton/exosuit to the human body, physical attachments also provide haptic – kinesthetic and proprioceptive – information to human skin, soft tissues, and joints near the attachment site. This haptic (touch-related) information is uncontrolled by the exoskeleton/exosuit and may have unintentional effects on human control of their own movement and muscle activation. The sensorimotor effects of physical attachments have not been little studied, and greater scientific understanding is important for informing future designs of physical attachments for safe and effective exoskeletons/exosuits.

Objective

The student will characterize effects of physical attachments on human movement and muscle activation. This characterization will demonstrate how existing exoskeletons/exosuits may unintentionally influence human behavior through haptic feedback.

Methods

The student will conduct human-subject experiments on unimpaired young adults to measure how different physical attachment types and donning procedures affect human movement and muscle activation patterns. The student will operate motion capture camera and EMG sensor systems to collect data on movement and muscle activity, respectively. Finally, the student will analyze this data in Matlab to extract biomechanics metrics such as joint range of movement and muscle activation level.

Prerequisites

Experience in or willingness to learn human-subject testing Matlab programming for data analysis Experience or interest in studying human biomechanics and sensorimotor control

Contact person

Mengnan.wu@vub.be

Promotor, co-promotor, advisor : michel.kinnaert@ulb.be, - , Louise Massager

Research Unit : SAAS

Description

Context

Europe is developing reusable launchers in the framework of different projects. To ensure launcher reliability, it is necessary to evaluate the state of health of the different parts of the launcher once it comes back to earth without dismantling it. The research project in which this master thesis takes place addresses the development of a systematic methodology to achieve this goal for the electromechanical actuators (EMAs) used to orientate the nozzle and the fins notably.

A model-based simulator of an electromechanical actuator (EMA) is currently used to complement the available experimental dataset (mostly in healthy operation) with synthetic data (of both healthy and faulty operating modes). There is however a mismatch between experimental data and synthetic data (i.e. generated by the simulator). Indeed, the simulator is based on simplified models of healthy operation and faults.

Objective

The goal is to improve the current model-based simulator to generate more realistic data (ultimately to better train the health monitoring system).

Methods

The main idea of the project is to achieve this goal by integrating a data-driven approach to the current simulator[1] in order to generate more realistic synthetic data (i.e. more similar to the limited experimental dataset we currently have and ideally to future experimental data). This way, future experimental data on natural degradation, defects and parameter variations (due to production variability, temperature change, etc.) could be exploited. The data-driven approach could for example be based on an adversarial machine learning approach [2], and on neural networks while enforcing adequate closed-loop performance [3].

The core challenges of the project lie in the limited available experimental dataset and in mitigating the impact of the data-driven approach. The latter point is essential for aerospace applications. Indeed the data-driven approach must not completely alter the physics-based models as the simulator must still be based on physics for ensuring explainability. This stems from the need for certification of the approach, which means that interpretability of how the method operates is required.

The work includes the following steps: 1. perform a bibliographic search on data-driven techniques for synthetic data generation, ideally for hybrid digital twins (i.e. digital twins based on both physics and experimental data). 2. study the EMA models in healthy and faulty states and get acquainted with the current simulator in MATLAB/Simulink 3. process the experimental measurements that will be provided to her/him in order to determine the most relevant features and/or identify parameters to update and relevant constraints 4. compare the most promising data-driven approach for this application in terms of performance and ease of updating once new data are available.

Prerequisite

*Good acquaintance with MATLAB/SIMULINK *Mastery of the bases of electrical machines and control theory

Contact person

For more information please contact : louise.massager@ulb.be and michel.kinnaert@ulb.be

references

[1] Massager, Louise, and Michel Kinnaert. "Modelling of electromechanical actuators in reusable launchers for health monitoring purposes." Internal report, 2024. [2] Goodfellow, Ian J., et al. "Generative adversarial nets." Advances in neural information processing systems 27 (2014). [3] Banderchuk, Ana, Daniel Coutinho, and Eduardo Camponogara. "Combining robust control and machine learning for uncertain nonlinear systems subject to persistent disturbances." 2023 62nd IEEE Conference on Decision and Control (CDC). IEEE, 2023.

attached pdf document

Promotor, co-promotor, advisor : michel.kinnaert@ulb.be, Laurent Catoire,

Research Unit : SAAS

Description

Context

Many of the pilot processes used in the SAAS department to teach control theory were developed through master's theses. This is the case for the rotary inverted pendulum, the ring positioner, the ball in the tube process, …

Objective

The aim of this master thesis is to develop new pilot processes that are modular, evolving, and open-source to provide a better and more practical learning experience to the students. Here are a few examples of processes that SAAS would like to develop (non-exhaustive list): - Ball in hoop or Flying ball in hoop o https://www.youtube.com/watch?v=8FaNk6C2ckM o https://www.youtube.com/watch?v=484GN4KBQnc o https://github.com/aa4cc/flying-ball-in-hoop o https://aa4cc.github.io/flying-ball-in-hoop/ - Cubli - robot that can jump up and balance on its corner o https://www.wevolver.com/specs/cubli - …

Methods

The main steps of the work are:

o selection of the sensors/actuators o design of the signal conditioning / acquisition stages o design of the experimental setup (SolidWorks, 3D printer …) o design of the power supply & cable management o modeling of the process o implementation of a control strategy (Arduino/C programming or Matlab/data-acquisition board or Raspberry PI) o setup of some didactic experiments & their related teaching materials

Prerequisite

• quick & autonomous learner in a dynamic environment
• team player, creativity
• basic knowledge in control theory, digital signal processing, electronics

Contact person

For more information please contact : laurent.catoire@ulb.be and michel.kinnaert@ulb.be

Reference

Xavier Jordens, Robin Wilmart, Emanuele Garone, Michel Kinnaert, Laurent Catoire. A Project-Based Learning Approach for Building an Affordable Control Teaching Lab: The Centrifugal Ring Positionner, IEEE Access, vol 10, pp 4907 – 4918, 2022

attached pdf document

Promotor, co-promotor, advisor : michel.kinnaert@ulb.be, - , Maxime Bussios

Research Unit : SAAS

Description

Context

Renewable energies and electric transportation are the cornerstones for developing a sustainable future society. Energy storage is fundamental in this context, in order to store surplus of energy and use it when the wind does not blow or the sun does not shine, or to produce vehicles that do not pollute the environment when they are on the roads. Among the possibilities, lithium-ion batteries are the technology of choice given their high energy capacity and efficiency. However in contrast with other battery technologies, the benefits of lithium-ion batteries come at the price of careful monitoring requirements. Indeed, faulty cells in a battery pack can have catastrophic consequences including fire.

Objective

The objective of this thesis is to develop a monitoring system that is able to detect and localize the degraded or weak cells within a pack on the basis of available voltage, current and temperature measurements. Both synthetic data obtained from a realistic battery pack simulator, and real data recorded on a 4-cell battery pack will be exploited to determine features that can be extracted from the measurements, or from combinations of measurements, and that exhibit pack malfunction. Next, appropriate classification tools will be investigated in order to decide on the healthy or degraded state of the pack and to localize the degraded cell/cells by processing the features extracted from the measurements. Various degradation levels and types will be considered in order to characterize the sensitivity to each fault.

Methods

The work can be separated in the following tasks: 1. perform a bibliographic search on fault/degradation diagnosis for battery packs, 2. generate synthetic data for heathy pack operation and for various degradation types and levels, 3. Use measurements and/or appropriate functions of the measurements to generate features that exhibit faulty/degraded behaviour, 4. Develop a classification method that decides on the pack state by processing the features extracted from regular measurements.

Requested skills

• quick & autonomous learner in a dynamic environment,
• team player, creativity,
• good knowledge of system and control theory,
• good acquaintance with MATLAB/SIMULINK

Contact person

For more information please contact : maxime.bussios@ulb.be and michel.kinnaert@ulb.be

attached pdf document

Promotor, co-promotor, advisor : michel.kinnaert@ulb.be, - , Louise Massager

Research Unit : SAAS

Description

Context

Europe is developing reusable launchers in the framework of different projects. To ensure launcher reliability, it is necessary to evaluate the state of health of the different parts of the launcher once it comes back to earth without dismantling it. This master thesis should contribute to a research project that addresses the development of a systematic methodology for the health monitoring of the electromechanical actuators (EMAs) used to orientate the nozzle and the fins notably. A classifier for detecting and identifying defects in an electromechanical actuator (EMA) is currently used based mostly on physical models of EMAs and common faults. The current classifier has several limitations which the present master thesis is aimed at overcoming.

Objective

The goal is to improve the current model-based fault detection and isolation (FDI) system. While additional ideas are welcome, different sub-goals have already been identified: • Automated adaptation to new data by incremental learning (no need for re-training from the complete database) • Automated definition of a new type of fault (i.e. generation of a new class) • Provide explanations on the decision taken (i.e. choice of classification; likelihood or belief associated to each class) • Account for the unbalanced database (many more data in healthy mode than in the different faulty modes) An existing classifier based on support vector machines is available[1] and the possibility to equip this type of classifier with the above properties should be evaluated. Possibly other types of classifiers could also be investigated if they offer more flexibility for achieving the indicated goals. Work on this topic has been ongoing for quite some time, notably for image processing[3] [4] [5]. In the present case, it is also important to come up with a solution with low computing cost and to provide an explanation for the decision taken[2]. Those constraints are essential as the approach is aimed for aerospace applications. It must therefore guarantee the performance of the classifier even after the inclusion of new data. The need for certification of the approach should be kept in mind, which means that interpretability of how the method operates is required.

Methods

The project is made of the following steps: 1. perform a bibliographic search on classifiers with adaptability to new data (incremental learning ability) and/or explainability features. 2. study the EMA models in healthy and faulty states and get acquainted with the current simulator in MATLAB/Simulink and FDI system [6]. 3. Get acquainted with the methodology to generate fault indicators and the associated features 4. Compare the most promising data-driven approach for this application in terms of performance and learning ability

Requested skills

• Quick & autonomous learner in a dynamic environment,
• Team player, creativity
• Good acquaintance with MATLAB/SIMULINK
• Good mastery of the bases in in control theory and digital signal processing

Contact person

For more information please contact : louise.massager@ulb.be and michel.kinnaert@ulb.be

references

[1] Wauthion, Benjamin, et al. "Monitoring based on analytical redundancy and classification for a primary flight surface electromechanical actuator." IFAC-PapersOnLine 55.6 (2022): 790-796 [2] Carlevaro, Alberto, et al. "Probabilistic safety regions via finite families of scalable classifiers." arXiv preprint arXiv:2309.04627 (2023). [3] Michael D. Muhlbaier, Apostolos Topalis, and Robi Polikar, Learn++.NC: Combining Ensemble of Classifiers With Dynamically Weighted Consult-and-Vote for Efficient Incremental Learning of New Classes, IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 20, NO. 1 (2009). [4] Liu Yu, Sarah Parisot, Gregory Slabaugh, Xu Jia and Ales Leonardis and Tinne Tuytelaars. More Classifiers, Less Forgetting: A Generic Multi-classifier Paradigm for Incremental Learning, European Conference on Computer Vision (ECCV), 2020 [5] Da-Wei Zhou , Yang Yang and De-Chuan Zhan, Learning to Classify With Incremental New Class, IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, VOL. 33, NO. 6 (2022). [6] Massager, Louise, Geoffrey Postal, and Michel Kinnaert. "Three-phase motor drive fault detection and isolation based on multi-model dual extended Kalman filtering." Benelux Meeting on Systems and Control, 2025.

attached pdf document

Promotor, co-promotor, advisor : benoit.scheid@ulb.be, - , Adam.Chafaï@ulb.be

Research Unit : TIPS

Description

For the production of therapeutic proteins in the posterior segment of the eyeball, gene therapy is the theoretically ideal option. The gene encoding the protein is introduced locally into cells which become endogenous internal bioreactors capable of producing it in situ [1]. The injection operation is shown in [2]. This operation is extremely delicate and requires automating the injection of the liquid containing the viral vectors. To do this, it is a question of optimizing this injection in particular to avoid damaging the cells of the retinal pigment epithelium (RPE) and avoid the reflux after injection. The objective of the master thesis will be to model the injection under artificial retina in a set-up built in the lab (already existing) and to identify the best pressure-control injection method to allow ophtalmologists to improve the surgical protocol. The master thesis will be performed in collaboration with ophtalmologists and possible test on animals (pigs). The work will be essentially experimental, but with some mathematical and numerical modeling of the process for understanding and optimization purposes. One of the challenge is to properly model (experimentally and theoretically) the adhesion force of the retina on the substrate.

For more information please contact : benoit.scheid@ulb.be

references

[1] https://www.em-consulte.com/em/SFO/2016/html/file100018.html [2] https://issuu.com/aoq-/docs/24959-aoq-revue-marsavril_2023-web

Promotor, co-promotor, advisor : benoit.scheid@ulb.be, - ,

Research Unit : TIPS

Description

Emulsification is a critical process in pharmaceutical manufacturing, especially in the context of vaccine formulation. The manufacturing environment poses considerable challenges, necessitating rigorous conditions to ensure product integrity. This project proposes to investigate enhancements to the vaccine formulation process within the industry by employing process intensification techniques. The master thesis will be realized in collaboration with GSK vaccines (Rixensart). The research will be two-pronged: it will involve experimental work to elucidate the dynamics of emulsification within mixing apparatuses, and theoretical analysis to explore the fundamental principles underpinning the process. To achieve this, we will leverage modeling tools, with a particular emphasis on Computational Fluid Dynamics (CFD). This dual approach aims to optimize the emulsification process, thereby improving the quality and efficacy of vaccines while potentially reducing production costs and environmental footprint. The master student will provide an invaluable platform to contribute to impactful research that could significantly advance public health outcomes. Contacts: Benoit Scheid (Benoit.Scheid@ulb.be)

Promotor, co-promotor, advisor : benoit.scheid@ulb.be, - ,

Research Unit : TIPS

Description

The project aims to develop a protocol for microencapsulating organoids in gel matrices in order to standardize their production and study the influence of matrix viscoelasticity on the development of a large number of organoids (healthy and tumorous). The mechanical properties of a specific set of encapsulated organoids will be studied in situ by applying mechanical deformations under living conditions. Organoids currently represent a recognized alternative to animal experimentation. The standardization of studies on a large number of organoids (several thousand) through the use of microfluidic tools should therefore have a significant impact on the industrial sector, which supplies research laboratories in particular. This work will be in collaboration with the « Institut de Recherche Interdisciplinaire en Biologie Humaine et Moléculaire (IRIBHM Jacques E. Dumont)” in the faculty of medicine at Erasme. contact person: Benoit Scheid (benoit.scheid@ulb.be)

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, - , Adam CHAFAI adam.chafai@ulb.be

Research Unit : TIPS

Description

Context: Percipio Robotics is a spin-off from the FEMTO-ST research institute, which has designed the Tulip gripper [1]. This compact, lightweight gripper, weighing less than 30g, is designed for micromanipulation and can grip objects from 50µm to 10mm. It solves the problems of large grippers and fragility frequently encountered in micro-robotics. Parallely, the TIPs department designs and manufactures compliant mechanisms in glass (FemtoPRINT technique), whose deformation is measured with optical/photonics techniques.

Objectives: This thesis aims to design and develop an instrumented version of the Percipio Robotics’ Tulip gripper. The master thesis can be preceded by a 3 months internship in the company (Besançon, France).

Methods: Literature review. Functional analysis and requirements. Design. Fabrication and characterization of the flexure mechanism.

Prerequisites: mechanical design, good command of French

Contact: adam.chafai@ulb.be

References: [1] https://ephj.ch/en/percipio-robotics-tulip-gripper-takes-micro-manipulation-to-the-next-level/ [2] L. Amez-Droz et Al. Instrumented Flexible Glass Structure: A Bragg Grating Inscribed with Femtosecond Laser Used as a Bending Sensor, MDPI Sensors, 23, 8018 (2023) [3] M. Tunon de Lara, Femtosecond pulse laser-engineered glass flexible structures instru-mented with an in-built Bragg grating sensor, Optics Express, https://opg.optica.org/oe/fulltext.cfm?uri=oe-31-18-29730&id=536683 (2023)

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, - , Manon CASSIGNOL (manon.nicole.francoise.cassignol@ulb.be)

Research Unit : TIPS

Description

Context: Soft matter can serve as an actuator in microrobotics by deforming under external stimuli such as light, heat, or pH, producing mechanical outputs like force or displacement. At the microscale, these smart materials can be 3D printed without assembly. In our lab, we use two-photon polymerization (2PP) to fabricate soft actuators from a thermo-responsive pol-ymer, poly(N-isopropylacrylamide) (pNIPAM). This material swells below its lower critical solution temperature (LCST) by absorbing water and shrinks above the LCST by expelling it. Recently, we fabricated 50 µm × 50 µm × 50 µm active cubes capable of bending, contract-ing, twisting, or shearing in heated water [1]. To achieve precise, multidirectional motion control, multiple actuators could be combined and selectively triggered by different wave-lengths of light. This is possible by doping them with photothermal nanomaterials that lo-cally convert light into heat [2]. Metallic nanostructures like gold (Au) and silver (Ag) nano-particles or nanorods have been used to actuate PNIPAM-based hydrogels [3]. However, they are usually dispersed uniformly, preventing spatial control. An alternative approach uses a tightly focused femtosecond laser in a PNIPAM hydrogel swollen with silver nitrate, locally forming Ag nanoparticles by multiphoton reduction [4]. Applying this method to our actuators would enable spatially selective nanoparticle patterning, allowing localized, pre-cise activation. Objective: The aim of this thesis is 3D print photosensitive nanoparticles (from a silver nitrate solu-tion) inside PNIPAM hydrogels with the 2PP machine. After printing, light will be used to il-luminate the actuators and will be converted into heat by the nanoparticles. The generated heat will trigger actuator motion by shrinking the hydrogel.

Methods: Literature review. Hydrogel fabrication (with 2PP printing or UV light). 2PP printing of Ag nanoparticles i.e., tune the printing parameters to obtain nanoparticles and optimize the actuation and mechanical properties, print complex deformation structures. Characteriza-tion: absorbance spectra, imaging the nanoparticles, and measuring the light responsive-ness of the structures.

Prerequisites: Materials (to develop the fabrication process and understand the behavior of the hydrogels with and without nanoparticles).

References: [1] Decroly, Gilles, Adam Chafaï, Guillaume de Timary, Gabriele Gandolfo, Alain Delchambre, et Pierre Lambert. 2023. « A Voxel‐Based Approach for the Generation of Advanced Kine-matics at the Microscale ». Advanced Intelligent Systems. [2] Cui, Ximin, Qifeng Ruan, Xiaolu Zhuo, Xinyue Xia, Jingtian Hu, Runfang Fu, Yang Li, Jianfang Wang, et Hongxing Xu. 2023. « Photothermal Nanomaterials: A Powerful Light-to-Heat Con-verter ». Chemical Reviews 123 (11): 6891 6952. [3] Park, Daehwan, Jin Woong Kim, et Chinedum O Osuji. 2024. « Programmable Thermo- and Light-Responsive Hydrogel Actuator Reinforced with Bacterial Cellulose ». [4] Nishiyama, Hiroaki, Shun Odashima, et Suguru Asoh. 2020. « Femtosecond Laser Writing of Plasmonic Nanoparticles inside PNIPAM Microgels for Light-Driven 3D Soft Actuators ». Op-tics Express 28 (18): 26470 80.

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, - , Manon CASSIGNOL (manon.nicole.francoise.cassignol@ulb.be)

Research Unit : TIPS

Description

Context: Soft matter is used as an actuator in microrobotics. It can deform under an external stimulus such as light, heat, or pH to generate a mechanical output (force and displace-ment). At the microscale, these smart materials can be 3D printed without assembly. In the lab, we use the two-photon polymerization method (2PP) to shape 50µm soft actuators out of a thermo-responsive polymer (pNipam = poly(N-isopropylacrylamide)). These active cu-bes demonstrate bending, contraction, twist, or shear deformation in a heated water bath [1]. Their mechanical performances such as Young modulus, force-displacement character-istics, or response time must now be characterized.

Objective: The aim of this thesis is to develop a setup to measure the force-displacement character-istics of such actuators. Inspired by Micro-Electro-Mechanical Systems (MEMS) force sen-sors [2] and/or atomic force microscopy (AFM) [3], this set-up will be fabricated in using glass microstructures (to be produced with the FemtoPrint machine) or with other materials deemed relevant by the candidate.

Methods: Literature review on characterizing the mechanical performance of soft material at mi-croscale. Select the most suitable device. Design the set-up considering the following crite-ria: 1) samples are characterized in water to allow them to swell and shrink, 2) a heating system (conventional or laser) will be used to drive the actuators, and 3) the sensor must be in contact with small samples (50 to 200 µm). Eventually, the results obtained may be supplemented and compared with data obtained with an environmental AFM, at UMons, and/or a nanoindentation system [4], at EMPA (Thun, Switzerland).

Prerequisites: Mechanics (to determine the device shape and develop the different part of the set-up using CAD software), coding (to automatically control the setup), and materials (to un-derstand the material model obtained from experimental measurements).

References: [1] G. Decroly, A. Chafaï, G. de Timary, G. Gandolfo, A. Delchambre, et P. Lambert, « A Voxel‐Based Approach for the Generation of Advanced Kinematics at the Microscale », Ad-vanced Intelligent Systems, 2023, doi: 10.1002/aisy.202200394. [2] M. Lamba, N. Mittal, K. Singh, et H. Chaudhary, « Design analysis of polysilicon piezoresis-tors PDMS (Polydimethylsiloxane) microcantilever based MEMS Force sensor », Int. J. Mod. Phys. B, vol. 34, no 09, p. 2050072, avr. 2020, doi: 10.1142/S0217979220500721. [3] A. Chau, S. Régnier, A. Delchambre, et P. Lambert, « Theoretical and Experimental Study of the Influence of AFM Tip Geometry and Orientation on Capillary Force », Journal of Adhe-sion Science and Technology, vol. 24, no 15 16, p. 2499 2510, janv. 2010, doi: 10.1163/016942410X508307. [4] T. Spratte et al., « Increasing the Efficiency of Thermoresponsive Actuation at the Mi-croscale by Direct Laser Writing of pNIPAM », Advanced Materials Technologies, vol. 8, no 1, p. 2200714, 2023, doi: 10.1002/admt.202200714.

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, Benoit HAUT, Clément RIGAUT

Research Unit : TIPS

Description

Context: The nose is responsible for heating and humidifying the air entering the respiratory tract. While it is only around 10 cm long, it can bring ambient air to a temperature of about 30°C in the pharynx. This function of conditioning the air before reaching the lower respiratory tract is vital to avoid inflammation, asthma and increased risk of infections. Despite the im-portance of this function of the nose, the heating of the air in the nasal cavity remains largely unknown.

Objective: This thesis aims to compute the temperature of the air exiting the nose under various conditions (rest, light effort, moderate effort,...) and ambient temperatures. The condition-ing efficiency of different noses can be compared to deduce the influence of anatomical features on air conditioning.

Methods: First, the simulation models will be created from STL files of nasal cavities. Then simula-tions will be carried out using OpenFOAM software to measure the temperature of the air exiting the nose for various parameters. Finally, the results of the different anatomies will be compared to extract the anatomical characteristics impacting air conditioning.

Prerequisites: • Fluid Dynamics • Thermodynamics

Contact: Clément Rigaut (clement.rigaut@ulb.be)

References: [1] D.-W. Kim, S.-K. Chung, et Y. Na, « Numerical study on the air conditioning characteristics of the human nasal cavity », Computers in Biology and Medicine, vol. 86, p. 18 30, juill. 2017, doi: 10.1016/j.compbiomed.2017.04.018. [2] S. Naftali, M. Rosenfeld, M. Wolf, et D. Elad, « The Air-Conditioning Capacity of the Human Nose », Ann Biomed Eng, vol. 33, nᵒ 4, p. 545 553, avr. 2005, doi: 10.1007/s10439-005-2513-4.  

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, Benoît HAUT, Clément RIGAUT

Research Unit : TIPS

Description

Context: While nasal drugs have been widely used to treat local symptoms of colds or allergies, they have more recently emerged as a potential method for delivering neurological drugs. Indeed, the nose is highly vascularized, which ensures that molecules deposited in the nasal cavity will be readily absorbed into the bloodstream. Moreover, there is growing evidence that drugs can pass directly from the nose to the brain via the olfactory nerve [1]. However, a major drawback of nasal administration is its strong dependence on the individual’s anatomy [2]. Therefore, personalized models are needed to predict the outcomes of nose-to-brain treatments.

Objective: This thesis aims to develop a pharmacokinetic model for the nasal administration of a neurological drug. The model will take into account the location of drug deposition (which varies between individuals), the transfer of the drug to the blood and brain, and its subse-quent elimination. The goal is to determine the optimal treatment plan for each patient (e.g., one large dose or multiple smaller doses, liquid or dry spray, etc.) and predict the in-dividual outcomes of these treatments.

Methods: Based on an extensive literature review, a pharmacokinetic model of the drug will be developed. Existing 3D-printed nasal replicas will be used to assess the distribution of the spray within the nasal cavity. These experimental data will allow for predictions of the treatment outcomes for a given anatomy and help identify the most suitable therapeutic approach for each individual.

Prerequisites: • Knowledge of a programming language

Contact: Clément Rigaut (clement.rigaut@ulb.be)

References: [1] L. Illum, ‘Is nose-to-brain transport of drugs in man a reality?’, Journal of Pharmacy and Pharma-cology, vol. 56, no. 1, pp. 3–17, Jan. 2004, doi: 10.1211/0022357022539. [2] C. Rigaut et al., ‘What Are the Key Anatomical Features for the Success of Nose-to-Brain Delivery? A Study of Powder Deposition in 3D-Printed Nasal Casts’, Pharmaceutics, vol. 15, no. 12, p. 2661, Nov. 2023, doi: 10.3390/pharmaceutics15122661.  

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, - , Margaux MANNNAERTS

Research Unit : TIPS

Description

Context: Lung cancer is the leading cause of cancer death worldwide [1]. As part of the screening process, lung nodules (suspected cancer) are regularly found in peripheral areas that are difficult to access by endoscopy. As most of these nodules are not cancerous, it is essential to be able to take a local biopsy to make a precise diagnosis. However, the lung is like a lab-yrinth, with sections that shrink with each division, and access to a precise peripheral zone is difficult. In addition, the need to use flexible and miniaturized tools implies certain limita-tions. Indeed, the need for flexibility is necessary to avoid damaging the tissue or injuring the patient but means that the tools may deform before the biopsy is taken.

A family of solutions that are being developed uses the concept of controllable/variable stiffness to cope with these issues [2]. These solutions use materials and/or specific geometries that can change rigidity given a certain stimuli (change of temperature, pressure, …).

Objectives: Develop a prototype of a variable stiffness catheter using different equipment present in the lab (molding techniques, 3D printers).

Methods: Literature review. Functional analysis and requirements. Design. Fabrication and evalua-tion of the built prototype.

Prerequisites: • Mechanical design • Interest for mechanical and biomedical engineering

Contact: margaux.mannaerts@ulb.be

References: [1] Global Burden of Disease 2019 Cancer Collaboration et al., « Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disability, and Disability-Adjusted Life Years for 29 Can-cer Groups From 2010 to 2019: A Systematic Analysis for the Global Burden of Disease Study 2019 », JAMA Oncol., vol. 8, no 3, p. 420, mars 2022, doi: 10.1001/jamaoncol.2021.6987. [2] L. Blanc, A. Delchambre, et P. Lambert, « Flexible Medical Devices: Review of Controllable Stiffness Solutions », Actuators, vol. 6, no 3, p. 23, juill. 2017, doi: 10.3390/act6030023.

attached pdf document

Promotor, co-promotor, advisor : pierre.lambert@ulb.be, - , Margaux MANNNAERTS

Research Unit : TIPS

Description

Context: Lung cancer is the leading cause of cancer death worldwide [1]. As part of the screening process, lung nodules (suspected cancer) are regularly found in peripheral areas that are difficult to access by endoscopy. As most of these nodules are not cancerous, it is essential to be able to take a local biopsy to make a precise diagnosis. However, the lung is like a lab-yrinth, with sections that shrink with each division, and access to a precise peripheral zone is difficult. In addition, the need to use flexible and miniaturised tools implies certain limita-tions. Indeed, the need for flexibility is necessary to avoid damaging the tissue or injuring the patient, but means that the tools may deform before the biopsy is taken. One way to ensure that the biopsy is taken at the right location is to have knowledge on the position and deformation of the catheter tip. Despite the exploration of various technologies such as electromagnetic sensors (EM), optical fibers, X-rays, etc [2], [3] , biopsy outcomes remain highly variable and dependent on a variety of factors including the type and number of used equipment, experience of the practician, location of the nodule in the lung. [4]

Objectives: This master thesis aims to design and develop a system enabling the practicians to know how the tip of the catheter is deformed in the lungs, due to their mechanical contact with the bronchii and the internal efforts developed in the catheter. Given the very small size of the peripheral bronchi (<1 mm), the system can be initially developed at a larger scale. Some inspiration can be taken from textile-based sensors, or other resistive strain gauges [5].

Methods: Literature review. Functional analysis and requirements. Design. Fabrication and charac-terization of a shape sensing catheter tip.

Prerequisites: • Mechanical design, electronics • Interest for mechanical and biomedical engineering

Contact: margaux.mannaerts@ulb.be

References: [1] J. M. Kocarnik et al., “Cancer Incidence, Mortality, Years of Life Lost, Years Lived With Disabil-ity, and Disability-Adjusted Life Years for 29 Cancer Groups From 2010 to 2019 A Systematic Analysis for the Global Burden of Disease Study 2019,” JAMA Oncol, vol. 8, no. 3, pp. 420–444, 2022, doi: 10.1001/jamaoncol.2021.6987. [2] C. Shi et al., “Shape sensing techniques for continuum robots in minimally invasive surgery: A survey,” IEEE Trans Biomed Eng, vol. 64, no. 8, pp. 1665–1678, Aug. 2017, doi: 10.1109/TBME.2016.2622361. [3] R. Brekken et al., “Accuracy of instrument tip position using fiber optic shape sensing for navigated bronchoscopy,” Med Eng Phys, vol. 125, Mar. 2024, doi: 10.1016/j.medengphy.2024.104116 [4] J. Thiboutot et al., “Accuracy of Pulmonary Nodule Sampling Using Robotic Assisted Bron-choscopy with Shape Sensing, Fluoroscopy, and Radial Endobronchial Ultrasound (The AC-CURACY Study),” Respiration, vol. 101, no. 5, pp. 485–493, Mar. 2022, doi: 10.1159/000522514. [5] S. Wu, « A n Overview of Hierarchical Design of Textile-Based Sensor in Wearable Electron-ics », Crystals, vol. 12, no 4, p. 555, avr. 2022, doi: 10.3390/cryst12040555.

attached pdf document

Promotor, co-promotor, advisor : benoit.haut@ulb.be, Pierre Lambert, Clément Rigaut

Research Unit : TRANSFERS, INTERFACES AND PROCESSES

Description

Context: The nose is responsible for heating and humidifying the air entering the respiratory tract. While it is only around 10 cm long, it can bring ambient air to a temperature of about 30°C in the pharynx. This function of conditioning the air before reaching the lower respiratory tract is vital to avoid inflammation, asthma and increased risk of infections. Despite the importance of this function of the nose, the heating of the air in the nasal cavity remains largely unknown.

Objective: This thesis aims to compute the temperature of the air exiting the nose under various conditions (rest, light effort, moderate effort,...) and ambient temperatures. The conditioning efficiency of different noses can be compared to deduce the influence of anatomical features on air conditioning. Methods: First, the simulation models will be created from STL files of nasal cavities. Then simulations will be carried out using OpenFOAM software to measure the temperature of the air exiting the nose for various parameters. Finally, the results of the different anatomies will be compared to extract the anatomical characteristics impacting air conditioning.

Prerequisites: • Fluid Dynamics • Thermodynamics

Contact: Clément Rigaut (clement.rigaut@ulb.be)

References: [1] D.-W. Kim, S.-K. Chung, et Y. Na, « Numerical study on the air conditioning characteristics of the human nasal cavity », Computers in Biology and Medicine, vol. 86, p. 18 30, juill. 2017, doi: 10.1016/j.compbiomed.2017.04.018. [2] S. Naftali, M. Rosenfeld, M. Wolf, et D. Elad, « The Air-Conditioning Capacity of the Human Nose », Ann Biomed Eng, vol. 33, nᵒ 4, p. 545 553, avr. 2005, doi: 10.1007/s10439-005-2513-4.

Promotor, co-promotor, advisor : benoit.haut@ulb.be, Pierre Lambert, Clément Rigaut

Research Unit : TRANSFERS, INTERFACES AND PROCESSES

Description

Context: While nasal drugs have been widely used to treat local symptoms of colds or allergies, they have more recently emerged as a potential method for delivering neurological drugs. Indeed, the nose is highly vascularized, which ensures that molecules deposited in the nasal cavity will be readily absorbed into the bloodstream. Moreover, there is growing evidence that drugs can pass directly from the nose to the brain via the olfactory nerve [1]. However, a major drawback of nasal administration is its strong dependence on the individual’s anatomy [2]. Therefore, personalized models are needed to predict the outcomes of nose-to-brain treatments.

Objective: This thesis aims to develop a pharmacokinetic model for the nasal administration of a neurological drug. The model will take into account the location of drug deposition (which varies between individuals), the transfer of the drug to the blood and brain, and its subsequent elimination. The goal is to determine the optimal treatment plan for each patient (e.g., one large dose or multiple smaller doses, liquid or dry spray, etc.) and predict the individual outcomes of these treatments.

Methods: Based on an extensive literature review, a pharmacokinetic model of the drug will be developed. Existing 3D-printed nasal replicas will be used to assess the distribution of the spray within the nasal cavity. These experimental data will allow for predictions of the treatment outcomes for a given anatomy and help identify the most suitable therapeutic approach for each individual.

Prerequisites: • Knowledge of a programming language

Contact: Clément Rigaut (clement.rigaut@ulb.be)

References: [1] L. Illum, ‘Is nose-to-brain transport of drugs in man a reality?’, Journal of Pharmacy and Pharmacology, vol. 56, no. 1, pp. 3–17, Jan. 2004, doi: 10.1211/0022357022539. [2] C. Rigaut et al., ‘What Are the Key Anatomical Features for the Success of Nose-to-Brain Delivery? A Study of Powder Deposition in 3D-Printed Nasal Casts’, Pharmaceutics, vol. 15, no. 12, p. 2661, Nov. 2023, doi: 10.3390/pharmaceutics15122661.

Promotor, co-promotor, advisor : benoit.haut@ulb.be, Jérémy Rabineau, Clément Rigaut

Research Unit : TRANSFERS, INTERFACES AND PROCESSES

Description

L'objectif de ce mémoire est de développer des outils et des procédures pour quantifier certains paramètres du flux de sang dans le ventricule gauche et l'aorte, en se basant sur des images d'IRM 4D flow.

En se basant sur des outils open-source existants, il s'agira de commencer par calculer quelques métriques basiques : volume d'éjection systolique, débit sanguin maximal, etc. Ensuite, des paramètres plus avancés, liés à la complexité du flux et à l'efficacité de l'activité cardiaque, seront intégrés : vorticité, hélicité, énergie cinétique turbulente, etc.

Ces outils pourront être testés sur des bases de données disponibles : comparaison d'astronautes avant et après vol spatial, comparaison de volontaires sains avant, pendant et après deux mois d'alitement "tête en bas" (simulation de micropesanteur), ou encore comparaison de patients et de volontaires sains, avec classification en fonction de certaines pathologies.

Contact : Benoit Haut

Promotor, co-promotor, advisor : john.lataire@vub.be, ICU director Prof. Joop Jonckheer, Andy Keymolen

Research Unit : VUB-ELEC

Description

Respiratory Oscillometry (RO) - Advancing Respiratory Monitoring: Refining Data-Driven Modelling for Longitudinal Assessment in Clinical Practice.

In this master's thesis project, you will delve into the realm of respiratory monitoring to unearth valuable insights that could improve patient care. The objective of this thesis is to discern statistically significant parameter variations across multiple measurements obtained from individual patients. The proposed approach is as follows: Getting up to topic: • Exploration of Respiratory Oscillometry (RO): Unravel the potential of RO in assessing lung impedance. Through a comprehensive survey of literature and state-of-the-art techniques, you will gain a deep understanding of how RO measurements can illuminate the health status of patients. • Hands-on Simulation and Mc Invent Trial Setup: Gain practical experience by simulating RO techniques in various realistic scenarios. Familiarize yourself with the intricacies of the clinical trial setup and the experiments conducted that gathered the data. • Data Processing Mastery: Learn the art of data cleaning and preprocessing to ensure the accuracy and reliability of measurements. Through meticulous removal of transients and breathing artefacts, you will prepare the data for in-depth analysis. Design and development of novel techniques. • Innovative Parameter Estimation Techniques: Elevate your skills by improving estimation techniques to enable longitudinal assessment of model parameters. Begin with a first-order model and progress towards more sophisticated models tailored to capture diagnostic-rich low-frequency regions [1]. • Efficient Automation: Streamline the analysis process by automating data processing, allowing for efficient examination of multiple patient datasets. • Development of User-Friendly Application: Translate your findings into actionable insights with the development of a user-friendly application. This application will provide clinicians with longitudinal assessments of identified parameters, from the MC Invent patients as an example but applicable in future trials, facilitating informed decision-making in patient care. • Collaboration with Physicians: Engage in discussions with physicians to validate the clinical relevance of identified parameters and ensure the ergonomic design of the application. Your collaboration will bridge the gap between technological innovation and clinical practice. • Predictive Modelling for Liberation from Mechanical Ventilation: Utilize the longitudinal assessments to develop a predictive model for determining a patient's readiness for weaning from mechanical ventilation. This predictive tool will empower clinicians to make informed decisions, ultimately leading to improved patient outcomes. Work approach: We encourage you to take ownership of the project, unleash your creativity, and push the boundaries of oscillometry. With access to dedicated workspace and state-of-the-art equipment, you will have the support and resources needed to excel in your research endeavors.

For more information please contact : John.Lataire@vub.be

References

[1] Bates, Jason HT. Lung mechanics: an inverse modelling approach. Cambridge University Press, 2009

attached pdf document

Promotor, co-promotor, advisor : john.lataire@vub.be, ICU director Prof. Joop Jonckheer, Andy Keymolen

Research Unit : VUB-ELEC

Description

Respiratory Oscillometry (RO) - Advancing Respiratory Mechanics Analysis: A time-varying modelling approach.

In the intricate dance of breathing, the inspiratory phase sees respiratory muscles contract, while during exhalation, they relax. This physiological ballet holds critical clues for clinicians, as contracted muscles typically exhibit increased stiffness. Leveraging this insight, we anticipate periodic fluctuations in the elastance estimate of a patient's respiratory system throughout each breath. For ICU physicians, this dynamic provides invaluable insights, with heightened inspiratory stiffness signalling strengthening respiratory muscles and potentially heralding readiness for liberation from mechanical ventilation. Yet, elastance is just the tip of the iceberg. As we delve deeper into within-breath analysis, we uncover a realm ripe for exploration within respiratory oscillometry. While non-parametric estimations have proven highly beneficial [1], parametric modelling has thus far shown promise primarily in simulation [2]. With this thesis, we endeavour to push the boundaries of innovation by crafting a robust parametric identification strategy for within-breath analysis in mechanically ventilated patients. Our overarching goal is clear: to discern statistically significant within-breath parameter changes across multiple measurements from individual patients. To achieve this, we propose a multifaceted approach:

Design and development of novel techniques. • Implement estimation techniques to detect a time-varying behaviour of the model parameters. Starting from [2] and improving towards more complex models • Design an excitation signal that allows an improved identification of the time-varying parameters • Design an experiment to prove that the excitation signal improves upon the state-of-the-art • Develop a longitudinal assessment strategy for the time-varying parametersWork approach: We encourage you to take ownership of the project, unleash your creativity, and push the boundaries of oscillometry. With access to a dedicated workspace and state-of-the-art equipment, you will have the support and resources needed to excel in your research endeavours.

For more information please contact : John.Lataire@vub.be

References

[1] Veneroni, Chiara, et al. "Oscillatory respiratory mechanics on the first day of life improves prediction of respiratory outcomes in extremely preterm newborns." Pediatric Research 85.3 (2019): 312-317. [2] Alamdari, Hamed Hanafi, Kamal El-Sankary, and Geoffrey N. Maksym. "Time-varying respiratory mechanics as a novel mechanism behind frequency dependence of impedance: a modeling approach." IEEE transactions on biomedical engineering 66.9 (2018): 2433-2446.

attached pdf document

Promotor, co-promotor, advisor : john.lataire@vub.be, ICU director Prof. Joop Jonckheer, Andy Keymolen

Research Unit : VUB-ELEC

Description

Respiratory Oscillometry (RO) - Investigating the Impact of Oscillometry on Ventilation Asynchrony: A Novel Approach

Ventilation asynchrony, a critical concern in patient care, occurs when ventilators fail to detect a patient's inspiratory effort in a timely manner. Traditionally, this effort is identified through peak flow signals generated during inspiration. However, the introduction of oscillometry into ventilation monitoring has introduced new challenges, as it overlays excitation signals onto the ventilation waveform, leading to subtle flow fluctuations during expiration. This master's thesis project aims to explore whether oscillometry exacerbates ventilation asynchrony. Leveraging patient data from the MC Invent trial, in this thesis you will analyze both oscillometry and ventilation measurements to discern any differences in ventilation asynchrony prevalence. Utilizing state-of-the-art identification techniques [1], the project will adapt methodologies to suit the specific context, enabling precise analysis. In scenarios where ventilation asynchrony is induced, the project will propose and validate an excitation signal design strategy aimed at mitigating its prevalence. This strategy will be tested on simulators to ensure efficacy and patient safety. Furthermore, in cases where ventilation asynchrony is not observed, the project will enhance the design of excitation signals. Current approaches often utilize exponential amplitude decay with a fixed peak-to-peak limit, which may not account for individual patient characteristics. By employing a model-based design strategy, the project seeks to optimize signal amplitudes and minimize flow perturbations that disturb the patient, thus improving patient comfort and outcomes.

Design and development of novel techniques. • Implement a ventilation asynchrony detection mechanism starting from [1] • Comparative analysis of ventilation asynchrony prevalence between ventilation and oscillometry measurements. • Development of an innovative excitation signal design strategy tailored to address ventilation asynchrony concerns or improve the amplitude design strategy. • Experimental validation of the proposed excitation signal's efficacy, surpassing existing benchmarks.

Work approach: To facilitate efficient collaboration and feedback, the project offers a dedicated workspace within the department. Access to a professional ventilator with oscillometry functionality enables real-world testing on lung emulators, ensuring the practicality and relevance of developed techniques.

For more information please contact : John.Lataire@vub.be

Reference

[1] L. van de Kamp, J. Reinders, B. Hunnekens et al., Automatic patient-ventilator asynchrony detection framework using objective asynchrony definitions. IFAC Journal of Systems and Control (2024), doi: https://doi.org/10.1016/j.ifacsc.2023.100236

attached pdf document

Promotor, co-promotor, advisor : john.lataire@vub.be, - , Andy Keymolen

Research Unit : VUB-ELEC

Description

Respiratory Oscillometry (RO) - Advancing Respiratory Monitoring: Extending Low-Frequency Oscillometry to Non-Invasive Mask Ventilation

In the Mc-Invent trial, low-frequency RO provided invaluable insights into respiratory mechanics of patients undergoing ventilation via an endotracheal tube (a tube inserted in the trachea). However, as patients transition to mask ventilation post-tube removal, the current monitoring techniques become ineffective, leaving clinicians without vital information crucial for patient care. Your mission: enhance low-frequency RO techniques to enable accurate estimations during mask ventilation. To achieve this, you will delve into the complexities posed by higher flow leakage and dead volume inherent in mask ventilation setups. Your approach will involve: In this thesis, you will improve the low-frequency RO technique such that viable estimations can be obtained during mask ventilation. You will adapt the RO techniques, the excitation signals and modelling processes, such that it can deal with the higher flow leakage (air that leaks via the mask boundaries) and the higher dead volume (air that is in between the sensor and the patient's mouth, encapsulated by the mask). Both elements influence respiratory mechanics estimations negatively. To achieve this, we suggest the following approach:

Design and development of novel techniques. • Understanding the Impact of Dead Volume and Leakage Flow: Through rigorous simulation studies, you will investigate the influence of dead volume and leakage flow on respiratory mechanics estimation. This foundational research will inform subsequent technique development. • Refinement of Modelling Strategies: Armed with insights from simulation studies, you will refine modelling strategies to mitigate the negative impact of dead volume and leakage flow. Incorporating prior knowledge, you will develop robust models capable of accurately capturing respiratory dynamics. • Innovative Excitation Signal Design: Central to your endeavor is the design of an excitation signal strategy that is resilient to the challenges posed by dead volume and leakage flow. Through novel design approaches, you will ensure the reliability and accuracy of respiratory mechanics estimations. • Experimental Validation: Rigorous experimentation is paramount to validate the efficacy of your developed techniques. You will design and conduct experiments to demonstrate the working principle and real-world applicability of your innovations, cementing their status as game-changing advancements in respiratory monitoring.

approach: We encourage you to take ownership of the project, unleash your creativity, and push the boundaries of oscillometry. With access to dedicated workspace and state-of-the-art equipment, you will have the support and resources needed to excel in your research endeavors

For more information please contact : John.Lataire@vub.be

attached pdf document