Comparaison de Problèmes Inverses pour la classification d'activité cérébrale en temps réel

Joan Fruitet 1 Maureen Clerc 1
1 ODYSSEE - Computer and biological vision
DI-ENS - Département d'informatique de l'École normale supérieure, CRISAM - Inria Sophia Antipolis - Méditerranée , ENS Paris - École normale supérieure - Paris, Inria Paris-Rocquencourt, ENPC - École des Ponts ParisTech
Abstract : This article compares the contribution of different methods of resolution of the inverse problem for the classification of brain activity in real time. A set of data from an somatotopic experiment in MEG has been used and an experiment of motor imaginary in EEG was performed. The aim is to classify the data in order to retrieve, from the MEG recordings, which finger has been stimulated, and for the EEG data, witch motor task was mentally performed. For this, various methods for solving the inverse problem are used to reconstruct the cortical activity and enhance the spatial resolution. A recursive algorithm using Student's t-test is proposed to select the cortical sources relevant for classification. The results are obtained using a Support Vector Machine (SVM). Overall, the use of the Minimum-Norm does not improve significantly the rate of classification. However, the use of the Beamforming increases the results from 80.2% to 83.4% for the MEG experiment and from 60% to 62% for the EEG experiment.
Document type :
Conference papers
Liste complète des métadonnées

Cited literature [9 references]  Display  Hide  Download

https://hal.inria.fr/hal-00727083
Contributor : Joan Fruitet <>
Submitted on : Saturday, September 1, 2012 - 1:29:36 PM
Last modification on : Friday, May 25, 2018 - 12:02:04 PM
Document(s) archivé(s) le : Sunday, December 2, 2012 - 3:28:13 AM

File

fruitet_260.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00727083, version 1

Collections

Citation

Joan Fruitet, Maureen Clerc. Comparaison de Problèmes Inverses pour la classification d'activité cérébrale en temps réel. GRETSI 2009, 2009, DIJON, France. ⟨hal-00727083⟩

Share

Metrics

Record views

443

Files downloads

207