2023-2024

Existing deep learning algorithms for predicting survival in Intensive Care Units gained attention in the recent years. However, their predictions are not reliable enough for decision support. This project intends to use test data to investigate where and why the current system gets its worse predictions, in order to propose
improvements.

• HealthTech Master student Rama Abdulhamid
• Master project supervised by Nicolas Lachiche et Vincent Castelain

Digital pathology has revolutionised the field of histological staining analysis since it provides a faster and more accurate diagnosis. Moreover, it has opened new opportunities for research and generated a strong demand for the development of Computer-Aided Diagnosis (CAD) systems. Although promising, this field introduces many challenges including the need for extensive time-consuming data annotations to cope with inter and intra-stain variabilities. To resolve this, we will develop an unsupervised deep learning method and demonstrate its capability to address multimodal variabilities without the manual overhead. Specifically, we will learn a one-to-many mapping by gradually translating the stain data from one modality to another using a transitive domain adaptation approach.

• HealthTech Master student : Ali Alhaj Abdo
• Master research project supervisors : Islem Mhiri, Thomas Lampert, Irene Spiridon, & Barbara Seeliger

• HealthTech Master student: Francesco Dettori
• Master research project supervisors: Stéphane Cotin & Florent Nageotte

This project is focused on parsimonious technologies and methods for upper limb rehabilitation. The need for low-cost technologies is a major challenge in the development of robot-assisted therapies for the rehabilitation of neuro-motor function disorders. This research started at ICube in 2022-23 in the framework of a joint project with Politecnico Milano, with two Master thesis projects, one focused on the development of a robotic exoskeleton structure, and the other on the design of bracelets designed to detect patient’s muscular activity onset. In the
present project, we aim at integrating the envisioned proofs of concept. This will require both to optimize the already proposed system and to design and control its actuation, in order to perform patient specific assistance.

• HealthTech Master student: Jesse Alvez
• Master research project supervisors: Bernard Bayel & Marta Gandolla

Active medical gestures guidance from pre-operative images is used in many medical fields, such as orthopedic surgery, for improving user’s accuracy. However, it is not used in general or digestive surgery notably because of pre-op to intra-op registration errors. The RDH team has developed techniques that learn such errors during the realization of the task and allow to adaptively guide the user. The objective of the project will be to develop and implement scenarios to assess under which cases guidance is possible and what are the conditions allowing improvement of the user’s performances.

• HealthTech Master student: Giorgia Baldazzi
• Master research project supervisors: Florent Nageotte & Barbara Seeliger

The arrival of private cellular technologies (ie. 5G) opens the possibility of equipping operating rooms with a single communication technology that will support both high bandwidth needs (video streams) and all patient monitoring
traffic during an operation. Medical operations of different natures suppose to have a robust and dynamic communication solution which will adapt to the different needs of information feedback while guaranteeing delivery times but also energy savings to preserve the lifetime of wireless monitoring devices.

• HealthTech Master student: Anna Beatriz de Souza Perotto
• Master research project supervisors: T. Noel & J. Montavont

Lewy Body Dementia (LBD) is a neuropsychiatric disorder and patients suffer from problems in mood, behavior and movement. Computer vision analysis of video recordings of body movement in LBD has problems to classify movement pathologies. The present project proposes to apply nonlinear recurrence analysis on the basis of multivariate time series. This new approach enables the extraction of complexity measures, which may classify pathological movement and distinguish it from healthy movement.

• HealthTech Master student: Gauthier Debes
• Master research project supervisors: Hyewon Seo & Axel Hutt

Almost 30% of patients with focal epilepsy are drug-resistant, leading to consideration of surgical resection of the epileptogenic zone. Intracranial exploration may then be required (stereo-electro-encephalography, SEEG),
involving the implantation of 10 to 15 depth electrodes to record electrical activity in the regions of interest. To limit risks (bleeding, infection), it is advisable to plan the most informative exploration possible while limiting the number of electrodes required. However, in current practice, manual planning is very time-consuming. The aim of this project is to develop a computer-assisted planning tool for SEEG enabling rapid decision-making. Based on interactive selection of regions of interest, the tool will provide trajectories that are parsimonious while maximizing sampling of the regions of interest. The planning tool will also offer interactive refinement of the proposed plan, with color-coded warnings based on fast evaluation of the constraints with each new configuration.

• HealthTech Master student: Léa Drolet-Roy
• Master research project supervisors: Caroline Essert, Irène Olliver & Lucas Gauer

The therapeutic strategy for cancer is driven by tumor aggressiveness. Two hallmarks of this aggressiveness are an altered metabolism, including that of glucose, and a lower pH. Magnetic resonance imaging (MRI), a non-irradiating modality, has recently been shown to allow in vivo measurement of the pH, although the long acquisition time impedes its adoption in clinical practice. The goal of this project is to accelerate the MRI acquisition so that it becomes usable for the evaluation of tumoral aggressiveness and patient follow-up in clinical practice.

• HealthTech Master student: Antoine Gilliet
• Master research project supervisor: Julien Lamy & Caroline Bund

Pancreatic cancer is a significant health concern, with intraductal papillary mucinous neoplasms (IPMNs) serving as early-stage precursors. The IMPULSE
project aims to enhance early detection through the integration of artificial intelligence and molecular profiling. The aim of this part of the project is to develop an original ensemble-based deep learning segmentation architecture for the automatic analysis of histopathological images to detect and quantify
on a massive scale these precancerous lesions.

• HealthTech Master student: Osman Gullu
• Master research project supervisor: Cédric Wemmert & Jérôme Cros

This project aims to consolidate an existing anatomical finite element model of the vestibule of the human inner ear, and to use it to improve the knowledge of the physiology of two common pathologies: endolymphatic hydrops and benign paroxysmal positional vertigo. To this end, various biomechanical models and numerical simulations are explored to lead to renewed diagnostic and therapeutic tools.

• HealthTech Master student: Mathias Halm
• Master research project supervisor: Daniel Baumgartner & Anne Charpiot

Femoral varus osteotomy is a surgical procedure aimed at correcting cases of knee malformation called genu valgum. During this operation, the femur is incised and reshaped using a plate to relieve stress on the affected side of the knee joint. However, there is a risk of post-operative fracture when
the patient starts bearing weight again. The first objective of this study is to develop a numerical model to quantify and evaluate the stress field in the bone and plate during the healing phase. The second objective is to establish and validate an experimental protocol to reproduce the deformations within the human femur during the post-operative phase.

• HealthTech Master student: João Luiz Machado Junior
• Master research project supervisor: Yamen Othmani, Massamaesso Bilasse, & Matthieu Ehlinger

Previous researches in medicine demonstrated that Alzheimer disease (AD) affects Hippocampus shape and the gait. In this thesis, we aim to investigate a joint analysis of 3D hippocampal shape and skeleton-based gait for AD diagnosis and monitoring. We propose to focus our first experiments on transgenic mice then we aim to transfer the retained results on human data. Thus we will perform domain adaptation to transfer the analysis on human data, in a translational approach.

• HealthTech Master student: Mubarak Olaoluwa
• Master research project supervisor: Hassen Drira & Laura Harsan

Glioblastomas are brain tumors that can be resistant to radiotherapy. The objective of this project is to develop new tools for analyzing multi-omics data in order to identify the factors contributing to radioresistance in high-grade glioblastomas. Deep learning models, specifically variational autoencoders (VAEs), will be employed to model this data. The novelty of this research lies in the use of a disentangled representation, previously used by the team for other applications. This disentangled representation will facilitate the identification and interpretation of common and discriminative signaling pathways associated with radioresistance and their impact on the response to irradiation.

• HealthTech Master student: Luis Villamarin
• Master research project supervisor: Julien Godet, Sylvain Faisan & Natacha Entz-Werlé

The aim of this multidisciplinary research project is to develop test-objects (=phantoms) offering cross-sectional representations of the human body, using materials with MRI-relaxometric properties similar to tissues. This type of phantom, also known as an anthropomorphic phantom, is quite well developed in CT, but is not widely available in MRI due to the complexity of obtaining materials with tissue-like relaxometric properties. MRI users (engineers, radiographers, radiologists, researchers) are waiting for such
devices to be able to test and program new sequences without having to mobilize a human being (patient, volunteer) over a long period of time (more than 1 hour of examination => ethical consideration). The aim of this work is to obtain MRI phantoms presenting both anatomical realism and contrast.

• HealthTech Master student: Habeeb Yusuff
• Master research project supervisor: Jean-Philippe Dillenseger, Simon Chatelin, & Pierre-Emmanuel Zorn