Project

Back to overview

Modeling pathological gait resulting from motor impairments: compare and combine neuromechanical simulation and machine learning approaches

Applicant Armand Stéphane
Number 177179
Funding scheme Sinergia
Research institution Laboratoire de Cinésiologie Willy Taillard Hôpitaux Universitaires de Genève
Institution of higher education University of Geneva - GE
Main discipline Interdisciplinary
Start/End 01.09.2018 - 28.02.2023
Approved amount 2'120'000.00
Show all

All Disciplines (4)

Discipline
Interdisciplinary
Neurophysiology and Brain Research
Information Technology
Biomedical Engineering

Keywords (6)

Machine learning; Neuromusculoskeletal modeling; Simulation; Gait; Motor impairments; Kinematics

Lay Summary (French)

Lead
La marche, essentielle pour les activités de la vie quotidienne de l’Homme, peut être altérée par des déficiences motrices (rétractions, faiblesses, spasticités, déficit de la sélectivité) dans le cas de nombreuses pathologies (ex. paralysie cérébrale, maladies neuromusculaires). Or, il est actuellement difficile de prédire l’impact de ces déficiences motrices sur la marche. Cette prédiction serait très utile pour aider au choix des traitements les plus appropriés.
Lay summary

L'objectif de ce projet est de modéliser et simuler la marche humaine dite « normale » et celle résultant de déficiences motrices. Pour relever ce défi, le projet adoptera une approche interdisciplinaire combinant et comparant un modèle neuromécanique de la marche et un modèle basé sur un apprentissage automatique. Ces deux modèles seront alimentés par une base de données conséquente d’individus ayant des déficiences motrices au niveau des membres inférieurs.

Ce projet apportera de nouvelles connaissances sur la locomotion humaine et les relations complexes entre les déficiences motrices et la marche. Il permettra de développer de nouveaux modèles, de nouvelles méthodes et de nouveaux outils pour la simulation et l'apprentissage automatique.

A terme, deux simulateurs seront accessibles gratuitement. Un simulateur éducatif de marche pathologique sera accessible au public pour simuler l'effet des déficiences motrices sur la marche. Un simulateur clinique sera également accessible aux équipes médicales afin d'explorer des scénarios hypothétiques simulant l'effet de traitements virtuels de déficiences motrices sur la marche d’un patient donné.

Direct link to Lay Summary Last update: 05.03.2018

Responsible applicant and co-applicants

Employees

Project partner

Publications

Publication
Exploring the Contribution of Proprioceptive Reflexes to Balance Control in Perturbed Standing
Koelewijn Anne D., Ijspeert Auke J. (2020), Exploring the Contribution of Proprioceptive Reflexes to Balance Control in Perturbed Standing, in Frontiers in Bioengineering and Biotechnology, 8, 1-12.
Impact of knee marker misplacement on gait kinematics of children with cerebral palsy using the Conventional Gait Model—A sensitivity study
Fonseca Mickael, Gasparutto Xavier, Leboeuf Fabien, Dumas Raphaël, Armand Stéphane (2020), Impact of knee marker misplacement on gait kinematics of children with cerebral palsy using the Conventional Gait Model—A sensitivity study, in PLOS ONE, 15(4), e0232064-e0232064.
{Data-Dependent} Conditional Priors for Unsupervised Learning of Multimodal Data
Lavda Frantzeska, Gregorová Magda, Kalousis Alexandros (2020), {Data-Dependent} Conditional Priors for Unsupervised Learning of Multimodal Data, in Entropy, 22(8), 888-888.
Improving {VAE} Generations of Multimodal Data Through {Data-Dependent} Conditional Priors
Lavda Frantzeska, Gregorová Magda, Kalousis Alexandros (2020), Improving {VAE} Generations of Multimodal Data Through {Data-Dependent} Conditional Priors, in {ECAI} 2020 - 24th European Conference on Artificial Intelligence, 29 August-8 September 2020, Santi, 325, 1254-1261, IOS Press, Portugal 325, 1254-1261.
Lower limb sagittal gait kinematics can be predicted based on walking speed, gender, age and BMI
Moissenet Florent, Leboeuf Fabien, Armand Stéphane (2019), Lower limb sagittal gait kinematics can be predicted based on walking speed, gender, age and BMI, in Scientific Reports, 9(1), 9510-9510.
Sample-Efficient Imitation Learning via Generative Adversarial Nets
Blonde Lionel, KalousisAlexandros (2019), Sample-Efficient Imitation Learning via Generative Adversarial Nets, in AI-STATS, 2019, Proceedings of Machine Learning Research, Naha, Okinawa, Japan.

Datasets

Marker misplacement of LKNE

Author Fonseca, Mickael; Armand, Stéphane
Publication date 21.10.2019
Persistent Identifier (PID) 10.26037/yareta:d7oanpxk3jd2xhd7sg7qhicdv4
Repository Yareta
Abstract
Data associated to a sensitivity study performed on misplacement of the lateral epicondyle marker (LKNE).- format .c3d and .csv

Collaboration

Group / person Country
Types of collaboration
LAMP, Centre National de Rééducation Fonctionnelle et de Réadaptation - Rehazenter Luxembourg (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
College of Health and Society, The University of Salford Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Industry/business/other use-inspired collaboration
Machine Learning and Data Analytics Lab, Friedrich-Alexander-Universität Erlangen-Nürnber Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Haute Ecole de Santé Vaud (HESAV) Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
- Exchange of personnel
Laboratoire de Biomécanique et Mécanique des Chocs (LBMC UMR_T9406) France (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Virtual - European Society of Movement Analysis for Adults and Children Poster Sensitivity of conventional gait model to lower limb marker misplacement 24.09.2020 Virtual Congress, Switzerland FONSECA Mickael; Armand Stéphane;
EPFL Bioengineering Day 2019 Poster Modeling Pathological Gait Resulting from Motor Impairments in Cerebral Palsy 27.11.2019 Lausanne, Switzerland Ijspeert Auke Jan; Di Russo Andrea; Dzeladini Florin; Stanev Dimitar;
European Society of Movement Analysis for Adults and Children Talk given at a conference Impact of lateral epicondyle marker misplacement on kinematics using the Conventional Gait Model – A sensitivity study 23.09.2019 Amsterdam, Netherlands Armand Stéphane; FONSECA Mickael;
9th International Symposium on Adaptive Motion of Animals and Machines Talk given at a conference Neuromechanical simulation of human locomotion: descending modulation of spinal reflex parameters during speed changes 20.08.2019 Lausanne, Switzerland Ijspeert Auke Jan; Di Russo Andrea;
9th International Symposium on Adaptive Motion of Animals and Machines Talk given at a conference Human Neuromuscular Balance Controller 20.08.2019 Lausanne, Switzerland Koelewijn Anne Daniëlle; Ijspeert Auke Jan;


Communication with the public

Communication Title Media Place Year
Media relations: print media, online media Un simulateur pour trouver le bon traitement SNF Profil German-speaking Switzerland Italian-speaking Switzerland Western Switzerland Rhaeto-Romanic Switzerland 2020
Media relations: print media, online media Marchez, vous êtes filmés Pulsations Western Switzerland 2019

Awards

Title Year
Motion Capture Ambassador Award 2020

Associated projects

Number Title Start Funding scheme
146801 Data-driven computer simulation of pathological gait resulting from contractures 01.09.2013 Project funding
197237 Controlling, planning, and learning movements in robots and animals 01.04.2021 Project funding
111111 Transforming Pius Branzeu Center of Laporoscopic Surgery and Microsurgery (PBCLSM) Timisoara into Eastern European Zonal Center of Development and Research in Laparascopic Surgery 01.01.2006 SCOPES
153299 Roombots: using reconfigurable robots to construct adaptive and assistive furniture 01.09.2014 Project funding

Abstract

Walking is often considered as the most important activity in daily living. The ability to move without pain, fatigue or any gait deviation is closely related to quality of life. However, many neuro-orthopaedic pathologies (e.g. cerebral palsy, stroke) involve impairments (i.e. weakness, reduced selective motor control, spasticity, soft tissue contractures, bone deformities and pain) that compromise normal movement. To choose the best treatment strategy it is of primary importance to understand the mechanisms of gait deviations and to identify the motor impairments that lead to them. Nevertheless, the links between motor impairments and gait deviations are very complex and lack scientific evidence. Moreover, there is no validated and easy-access simulation tool that permits simulating the influence of motor impairments and their treatments in patients with gait deviations. The present project aims to model and to simulate pathological gait resulting from motor impairments. To address this challenging aim, the project will take an interdisciplinary approach that combines a neuromechanical model and a data-driven model fed by a pathological gait data repository. More specifically, a data repository of healthy and pathological gait (>800 examinations) will be established and analyzed from a large existing database of the Kinesiology Laboratory as well as from new experiments designed to quantify intrinsic and extrinsic gait variability. The data repository will be used to build and to evaluate a neuromechanical model of gait (i.e. a computational model of the musculoskeletal system and of the spinal locomotor circuits) including motor impairments and to train a data-driven model using machine learning methods in order to predict a patient’s gait from their clinical data and to predict gait modifications from modifications of clinical data (virtual treatment). Moreover, a joint framework of machine learning and neuromechanical simulation for gait modeling will be built and tested where the one will support the other creating a virtuous cycle. Concretely, machine learning models will be used to support the simulation of the neuromechanical model; and neuromechanical simulations will be used to generate a large amount of data that will be used to train the data-driven model. The best models will be selected, based on the results of systematic evaluations and comparisons using appropriate performance metrics, for integration in two simulators. An educational pathological-gait simulator will simulate the effect of motor impairments on gait and a clinical pathological-gait simulator will provide the means to explore what-if scenarios simulating the effect of virtual treatments for a given patient. The project will build on the results of a previous SNF project, “Data-driven computer simulation of pathological gait resulting from contractures” (http://p3.snf.ch/project-146801) that aimed to emulate the effect of soft tissue contractions on the gait of healthy individuals. The results included among others an online educational simulator of gait produced in the presence of contractures (http://pgcs.unige.ch). In the present project, we want to go considerably further in the simulation and modeling of pathological gait exploiting a unique database of pathological gait and building on the expertise of a complementary interdisciplinary team. The team includes experts in clinical gait analysis and biomechanics (Dr Armand’s team), experts in neuromechanical modeling (Prof Ijspeert’s team), and experts in machine learning (Dr Kalousis’s team). Moreover, a clinician (Dr De Coulon, Orthopedic surgeon) specialist of neuro-orthopaedic pathologies in pediatric population (e.g. cerebral palsy) and a company expert in developing simulation tools (Cyberbotics) will complete the interdisciplinarity needed for this project.The project will bring new knowledge about human locomotion and the complex relationships between motor impairments and gait deviations, new models, new simulation methods and tools, and new machine learning method tailored to build gait models. Fundamental as well as clinical contributions are expected that will improve the treatments related to pathologies with motor impairments affecting gait (e.g. cerebral palsy patients).
-