Improved fMRI group studies based on spatially varying non-parametric BOLD signal modeling

Multi-subject analysis of functional Magnetic Resonance Imaging (fMRI) data relies on within-subject studies, which are usually conducted using a massively univariate approach. In this paper, we investigate the impact of a novel within-subject analysis on group studies. Our approach is based on the use of spatial mixture models (SMM) in a. joint detection-estimation framework (JDE) [1]. This setting allows us to characterise the hemodynamic filter at a regional scale and therefore to account for its spatial variability. As the subject- specific BOLD effects enter as input parameters in the computation of group statistics, we then compare two kinds of Random effect analyses (RFX). The first one takes the estimated BOLD effects computed by SPM¹ as inputs while the second one considers the results of our JDE scheme. We finally show on a real dataset of 15 subjects that brain activations appear more spatially resolved using SMM instead of SPM and that a better sensitivity is achieved. Moreover, the JDE framework allows to assess the regional inter-subject variability of the brain dynamics.