Functional parcellation of the human face-selective areas: a resting-state connectivity homogeneity analysis

Although different functional parcellations of human face-selective areas in the ventral visual stream have been proposed, the extent to which such regions share similar properties is still unclear1. Using a subset of 165 subjects from the HCP1200 functional magnetic resonance imaging dataset, we combined task-based definition of face-selective areas and resting-state connectivity homogeneity analysis to reach a finer-grained parcellation of face-selective areas. We segmented individual surface-based statistical maps of activation during vision of faces vs places through a watershed algorithm, identifying three parcels lying between the fusiform and the middle occipital gyrus. We then computed a connectivity fingerprint of each face-selective surface vertex, defined as the correlation between its resting-state timeseries and that of all vertices outside the face parcels. Finally, we computed a homogeneity measure between the connectivity fingerprints of each pair of vertices within and between the face parcels. The connectivity homogeneity between face vertices was inversely proportional to the anatomical distance between them measured on the individual cortical surface. Having accounted for distance effects, homogeneity was higher within than between face parcels. This result shows that three face regions are distinguishable based on their resting-state functional connectivity fingerprints. Despite needing validation, our method shows the possibility of segregating parcels with similar functional properties using the profile of their functional connectivity at rest.