A variational method to determine the most representative shape of a set of curves and its application to knee kinematic data for pathology classification [r-libre/1381]

Ben Nouma, Badreddine; Mezghani, Neila; Mitiche, Amar et Ouakrim, Youssef (2018). A variational method to determine the most representative shape of a set of curves and its application to knee kinematic data for pathology classification. Dans Proceedings of the second Mediterranean Conference on Pattern Recognition and Artificial Intelligence (MEDPRAI 2018). New York, USA : Association for Computing Machinery, coll. « ACM International Conference Proceedings Series ». ISBN 978-1-4503-5290-1 https://doi.org/10.1145/3177148.3180095

Fichier(s) associé(s) à ce document :

PDF - MedPrai2018_VF1.pdf
Contenu du fichier : Version de l'éditeur
Accès restreint

Demander une copie

Catégorie de document :	Communications dans des actes de congrès/colloques
Évaluation par un comité de lecture :	Oui
Étape de publication :	Publié
Résumé :	ABSTRACT The purpose of this study is to investigate a variational method to determine the most representative shape of a family of curves and its application to three-dimensional knee kinematic data for knee pathology classification. High variability and the presence of outliers are characteristic of the data in this application. This method determines the most representative shape by averaging the family curves corrected to account for outliers occurrence and family variability. To this effect, the correction is erformed by simultaneous minimization of a set of objective functions, one for each family curve onsisting of two terms: a data term of conformity of the corrected curve to the given family curve, and a regularization term of proximity of the corrected curve to the mean of the corrected curves to inhibit the influence of outliers in the family. Minimization is carried out efficiently by particle swarm optimization, a method which, in contrast to gradient descent, is robust to the presence of outliers. Experimental eesults using real-world data in knee osteoarthritis pattern classification demonstrate the validity and efficiency of the method. Comparisons to conventional methods used to determine the most representative shape are given. KEYWORDS Knee kinematic data curves, knee pathology classification, particle swarm optimization
Déposant:	Mezghani, Neila
Responsable :	Neila Mezghani
Dépôt :	02 févr. 2018 14:33
Dernière modification :	12 déc. 2019 14:16

Actions (connexion requise)

RÉVISER