A0379
Title: Generation of synthetic mixed data for multiple sclerosis patients: Application to gait data and EDSS score
Authors: Klervi Le Gall - Nantes University (France) [presenting]
Lise Bellanger - Nantes University (France)
Aymeric Stamm - CNRS - Laboratoire de Mathematiques Jean Leray (France)
David Laplaud - CR2TI -INSERM U1064 - CHU de Nantes- Nantes University (France)
Abstract: Gait analysis is a key factor in the understanding and care of multiple sclerosis. To analyse gait, we developed a biomarker which characterises the rotation of the hip of an individual during an average gait cycle using unit quaternion time series (QTS) collected with a motion sensor. To complete the data, EDSS scores (qualitative markers of disease progression) were assessed by neurologists. We developed a promising clustering method to group patients with similar gait impairments using a small database. We need a larger volume of data to assess its robustness and validate the method. We propose a sound statistical method to generate synthetic mixed data, including QTS and EDSS scores which best resembles the original data set while keeping anonymity and preserving inertia. To the best of our knowledge, this is the first time that such a method is proposed for functional data evaluation in the Lie group of 3D rotations. The proposed approach builds a synthetic data set by mixing multivariate functional PCA, multivariate analysis of mixed data and nearest neighbours weighting. We will show the relevance of our approach using a sample of multiple sclerosis patients from a clinical study conducted in collaboration with the University Hospital of Nantes.
