Impact of the joint detection-estimation approach on random effects group studies in FMRI

Inter-subject analysis of functional Magnetic Resonance Imaging (fMRI) data relies on single intra-subject studies, which are usually conducted using a massively univariate approach. In this paper, we investigate the impact of an improved intra-subject analysis on group studies. basically the joint detection-estimation (JDE) framework where an explicit characterization of the Hemodynamic Response Function (HRF) is performed at a regional scale and a stimulus-specific adaptive spatial correlation model enables the detection of activation clusters at voxel level. For the group statistics, we conducted several Random effect analyses (RFX) which relied either on the General Linear Model (GLM), or on the JDE analyses, or even on an intermediate approach named Spatially Adaptive GLM (SAGLM). Our comparative study perfomed during a fast-event related paradigm involves 18 subjects and illustrates the regionspecific differences between the GLM, SAGLM and JDE analyses in terms of statistical sensitivity. On different contrasts of interest, spatial regularization is shown to have a beneficial impact on the statistical sensitivity. Also, by studying the spatial variability of the HRF, we demonstrate that the JDE framework provides more robust detection performance in cognitive regions due to the higher hemodynamic variability in these areas.