Which registration method for high resolution fMRI to explore hand movement cortical representation?

Progress in our understanding of brain functions relies on our capability to explore the human cortical surface at a fine scale (typically 1.5 mm isotropic at 3T). For this purpose, high accuracy is required for all processing steps from image acquisition to data analysis. For group studies, the high intersubject variability of the human cortices hampers their precise registration. Based on the hypothesis that function follows anatomy, accurate inter-subject sulci registration should result in precise alignment of corresponding functional regions and then improve the statistical significance of parametric maps. Converging evidence from intra-operative mapping, cytoarchitectony, and functional neuroimaging conclusively rely primary cortical regions to morphological landmarks. Thus, the so-called "hand knob" landmark, a specific feature of the central sulcus (CS), consistently separates the primary motor cortex (M1), on its anterior bank, from the primary sensory cortex (S1) on its posterior bank. In an effort to define a dedicated processing pipeline for a fine non-invasive exploration of human M1, we compared four nonlinear registration methods applied on high resolution fMRI of basic hand movements. First, based on anatomical measures, we show how recent local or global diffeomorphic techniques improve the alignment of sulcal folds in M1. Second, with functional measures, we quantitatively evaluate their effect on the detection and localization of activation clusters at the population level. Based on such measures, we show that two diffeomorphic methods working globally (DARTEL) or including sulcal constraints (DISCODARTEL) improve activation detection and localization opening the way to a non-invasive exploration of the human hand motor cortex. © 2012 IEEE.