Improved intersubject colocalization of functional activations by nonlinear fitting of individual contrast maps to a functional atlas

A comparison of activation maps obtained with fMRI reveals substantial inter-individual variability in the anatomical location of activated areas. In group studies functional data are smoothed with a Gaussian kernel to increase the functional overlap between activated areas from different subjects. We present an alternative approach to increase this overlap, based on non-linear deformations of individual contrast maps to a sample-specific minimal deformation target. Functional images were obtained from six adult subjects passively viewing a short stimulus sequence. The stimuli were organized in a blocked design with three conditions: static objects, faces and moving natural scenes. All functional images were spatially normalized to the standard MNI template. Individual statistical maps were calculated, contrasting each of the three conditions with a fixation condition. Deformation fields resulting from viscous fluid registrations between these individual contrast maps were obtained to create sample-specific minimal deformation targets for the three contrasts. For each subject we then calculated the average deformation needed to register each contrast map with its associated target. This subject-specific deformation field was subsequently applied to all functional images for that specific subject. This procedure does increase the inter-subject overlap in activated areas, as evidenced by a comparison between the results of fixed-effects group analyses on the deformed and on the undeformed functional images. The analysis on the functionally deformed images yields about one third more activated clusters than the analysis on the functionally undeformed images (33 versus 24). With functionally deformed images the contrast between moving scenes and fixation reveals a number of areas that are not significantly activated with functionally undeformed images: hMT/V5, the frontal eye fields (FEF) and both the anterior and lateral dorsal intraparietal sulcus regions (DIPSA and DIPSL). We believe the enhanced functional overlap generated by this functional atlas fitting paradigm improves the analysis of functional group data. Moreover, this paradigm provides a means for a direct mapping of a functional brain atlas to individual contrast maps. As such, it may enable automated labeling of functional areas and as such provide a valuable diagnostic tool for functional patient studies.status: publishe