An a Contrario Approach for the Detection of Activated Brain Areas in FMRI

Camille Maumet¹, Pierre Maurel¹, Jean-Christophe Ferr¹, ², Christian Barillot¹

In this paper we propose an a contrario approach for the detection of activated brain areas in BOLD functional MRI data. The method focuses on subject-specific activations. Comparisons with the standard massively univariate general linear model are provided using Receiver-Operating-Characteristics curves for two different ground truths. The data used for validation include 12 subjects, scanned 3 times for a hand-motor paradigm. Overall, the proposed a contrario approach displayed better spatially defined activation with a more interesting trade-off between sensitivity and specificity by comparison to the standard massively univariate GLM.