A PhD position is available in IMT Atlantique, Brest campus, France, team RAMBO (https://www.imt-atlantique.fr/en/research-innovation/teams/rambo) in the following theme :
FIne-grained and cusTomized, Neuromusculoskeletal-based assESSment of rehabilitation exercises – FITNESS
– Domain and scientific/technical context
Project FITNESS will develop an improved computer vision-based approach for functional capacity evaluation (FCE), namely, the assessment of a person’s ability to perform daily living activities or work tasks and more specifically, the evaluation of exercises in the context of patient supervision who are undergoing rehabilitation following a surgery, an accident or for the treatment of a musculoskeletal disorder. Nowadays, assessment and patient progress are heavily reliant on clinical expertise which is costly in terms of time and human resources, while the latest artificial intelligence (AI)-based methods have demonstrated a capacity to attain human-level analysis performance [1], [2]. Still, to further integrate AI-based methods in the current practice of FCE, they need to be customized to an individual via a spatiotemporally localized analysis of the movement accounting for both visual (3D skeleton) as well as latent, neuromusculoskeletal features (i.e. muscle activations and forces).
– Scientific/technical challenges
Project FITNESS will build upon and extend state-of-the-art methods [1], [2] recently developed within the team, showing to outperform existing, machine-learning-based approaches in the assessment of physical rehabilitation exercises via the analysis of 3D human skeleton trajectories extracted from ordinary RGB videos. However, current methods are memoryless and agnostic to the body of each individual. This means that they assess a global quality score over a complete video recording without accounting for the particularities of the individual 3D skeleton under consideration while they cannot provide to the patient focused feedback with respect to the erroneously executed movement segments. Finally, they assess independently each video sequence which does not allow monitoring of the patient’s progress and they solely rely on visible information, namely, human movement, whereas the resulting movement execution is the result of multiple neuromusculoskeletal activations that are latent and cannot be captured by vision sensors.
– Considered methods, targeted results, and impacts
Project FITNESS will pursue interdisciplinary research by accounting both for visual features as well as latent neuromusculoskeletal features of body movement. To account for the latter, visually captured human movement will be further instantiated via biomechanical digital twins (DT) of the human through tools used in the medical domain, namely, OpenSIM (https://simtk.org/projects/opensim) and Hufydy (https://hyfydy.com/) that account for the biomechanical structure of the human body as well as physics constraints. By jointly exploiting skeleton as well as neuromusculoskeletal features in a joint machine-learning framework, we target the following main results/impacts: (i) enhancing the accuracy of functional capacity evaluation, (ii) allowing for fine-grained feedback to the user and the medical expert with respect to the interplay of muscles that are employed, (iii) enabling the generation of synthetic/simulated movements for enhanced training of machine-learning algorithms.
[1] Ikram Kourbane, Panagiotis Papadakis, Mihai Andries, SSL-Rehab: Assessment of physical rehabilitation exercises through self-supervised learning of 3D skeleton representations, Computer Vision and Image Understanding, 251, 2025, https://doi.org/10.1016/j.cviu.2024.104275
[2] Ikram Kourbane, Panagiotis Papadakis, Mihai Andries, Optimized Assessment of Physical Rehabilitation Exercises using Spatiotemporal, Sequential Graph-Convolutional Networks, Computers in Biology and Medicine, 186, 2025, https://doi.org/10.1016/j.compbiomed.2024.109578
For full details on the context and the candidature procedure see the link below :
