Classification de sujets asymptomatiques et gonarthrosiques en fonction des données cinématiques comparaison de l'approche globale et de l'approche locale

Abstract

This study is aimed at developing a way to automatically classify asymptomatic subjects (AS) and osteoarthritic subjects (OA) based on global and local kinematic data obtained during one normal walking cycle. Fourteen asymptomatic participants and 30 osteoarthritic patients (divided into two groups according to the severity of their disease, based on the Kellgren & Lawrence (KL) scale: 16 mild cases, and 14 advanced cases) took part in this study. The classification system developed consists of three modules: 1- a module for kinematic data acquisition (adductionabduction, flexion-extension, and internal-external rotation) by means of an exoskeleton placed over the knee joint; 2- a module for extracting the characteristics of the data using two approaches, global and local. The global approach characterizes all the kinematic data, seeking the most discriminating subcycle of the kinematic signal, while the local approach characterizes the kinematic data by points of interest; and 3- a module for data classification in two hierarchical levels, one to discriminate AS subjects from OA patients, and the other to sort the severity of the disease of the OA patients into two groups (OAKL1,2 and OAKL3,4). The classification system selected is based on a subspace method, using singular value decomposition (SVD). The classification results obtained show that the adduction-abduction kinematic signal is the most discriminating for both the global and the local approach. The rate of classification of the global representations of the AS subjects and OA patients is 93.18%, and 93.33% between OAKL1,2 and OAKL3,4. The rate of classification of the local representations between the AS subjects and the OA patients is 77.27%, and 83.33% between the OAKL1,2 and OAKL3,4 patients. In conclusion, a better hierarchical classification between the AS subjects and the OA patients, and between the degrees of severity of the disease in the OA patients (OAKL1,2 and OAKL3,4), is achieved with the global adduction-abduction kinematic data.

Date	24 Oct 2011
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Neila Mezghani (Supervisor) & Jacques A. de Guise (Co-supervisor)