Graph-based inter-subject classification of local fMRI patterns

Abstract : Classification of medical images in multi-subjects settings is a difficult challenge due to the variability that exists between individuals. Here we introduce a new graph-based framework designed to deal with inter-subject functional variability present in fMRI data. A graphical model is constructed to encode the functional, geometric and structural properties of local activation patterns. We then design a specific graph kernel, allowing to conduct SVM classification in graph space. Experiments conducted in an inter-subject classification task of patterns recorded in the auditory cortex show that it is the only approach to perform above chance level, among a wide range of tested methods.
Type de document :
Communication dans un congrès
Third International Workshop Machine Learning in Medical Imaging - MLMI 2012 (Held in Conjunction with MICCAI 2012), Oct 2012, Nice, France. Springer Berlin Heidelberg, 7588 (0302-9743), pp 184-192, 2012, Lecture Notes in Computer Science. 〈10.1007/978-3-642-35428-1_23〉
Liste complète des métadonnées

Littérature citée [15 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-00818815
Contributeur : Sylvain Takerkart <>
Soumis le : lundi 29 avril 2013 - 11:45:57
Dernière modification le : lundi 11 février 2019 - 11:12:03
Document(s) archivé(s) le : lundi 19 août 2013 - 09:47:20

Fichier

mlmi2012_takerkart_camera_read...
Fichiers produits par l'(les) auteur(s)

Identifiants

Citation

Sylvain Takerkart, Guillaume Auzias, Bertrand Thirion, Daniele Schön, Liva Ralaivola. Graph-based inter-subject classification of local fMRI patterns. Third International Workshop Machine Learning in Medical Imaging - MLMI 2012 (Held in Conjunction with MICCAI 2012), Oct 2012, Nice, France. Springer Berlin Heidelberg, 7588 (0302-9743), pp 184-192, 2012, Lecture Notes in Computer Science. 〈10.1007/978-3-642-35428-1_23〉. 〈hal-00818815〉

Partager

Métriques

Consultations de la notice

753

Téléchargements de fichiers

367