Utilisation d'informations géométriques pour l'analyse statistique des données d'IRM fonctionnelle

Guillaume Flandin 1
1 EPIDAURE - Medical imaging and robotics
CRISAM - Inria Sophia Antipolis - Méditerranée
Abstract : Functional magnetic resonance imaging (fMRI) is a recent modality allowing to measure in vivo the neuronal activity of healthy subjects or patients and thus to investigate the link between cerebral structure and functioll. We are interested in introducing cerebral anatomy to analyse functional data. We thus reconsidered the classical analysis usually performed voxel-by-voxel following a spatial smoothing to propose a representation of data relying on an anatomo-functional parcellation of the cortex. This representation allows to reduce the dimensionality of the data into a small number of elements, more relevant from a neuroscience point of view. We present several examples of application of this parcellation approach, at first based on anatomy only. An activity detection study based on a linear model highlights an increased sensitivity compared to the voxel-by-voxel approach. We also present two other applications using parcellations, dealing with the selection of regional models and functional connectivity studies. This description also allows to propose a solution to the problem of group analyses, where subjects may exhibit an important anatomo-functional variability. To bypass the difficult problem of registration between different subjects, we propose a parcellation approach grouping the homogeneous areas from an anatomical and functional point of view, between subjects. The application of this method on a functional proto col reveals that it actually can take into account the anatomo-functional variability and thus offers sorne robustness to multi-subject analyses.
Document type :
Complete list of metadatas

Contributor : Service Ist Inria Sophia Antipolis-Méditerranée / I3s <>
Submitted on : Tuesday, October 18, 2011 - 4:33:24 PM
Last modification on : Saturday, January 27, 2018 - 1:30:51 AM
Long-term archiving on : Thursday, January 19, 2012 - 2:30:14 AM


  • HAL Id : tel-00633520, version 1



Guillaume Flandin. Utilisation d'informations géométriques pour l'analyse statistique des données d'IRM fonctionnelle. Interface homme-machine [cs.HC]. Université Nice Sophia Antipolis, 2004. Français. ⟨NNT : 2004NICE4019⟩. ⟨tel-00633520⟩



Record views


Files downloads