Abstract
An ideal atlas of the human brain provides consistency across studies and populations, corresponds to meaningful units of cognitive processing, and captures individual differences. Functional connectivity is a promising candidate for defining atlases because it can capture individual differences and predict behavior, brain disorders, and cognition. While many individualized atlases based on functional connectivity have been proposed, there is no single correct solution. Further, it remains unclear how these atlases relate to an individual’s functional regions of interest (fROI), such as category-selective visual regions. Here, we describe a more flexible approach to determine individual solutions across a wide range of resolutions. To construct and validate these personalized atlases, we utilized a large sample of young adults (N= 1,018) with resting state and task fMRI from the Human Connectome Project (van Essen et al. 2012). Solutions based on voxel-wise connectomes were calculated in a two-step approach: first computing group solutions using k-means, then determining individual solutions based on a voxel’s similarity to the group network’s connectivity profiles. We found these personalized atlases: 1. Conserve group-level aspects, 2. Replicate broadscale organization of previously released atlases, and 3. Are stable within individuals. Next, we identified parcels from these resting state atlases (rsROIs) corresponding to fROIs for motion, working memory, and high-level vision. We further investigated the following category-selective visual regions: one body region (EBA), three face regions (FFA, OFA, STS), three place regions (PPA, RSC, TOS), and two tool regions (LO, OFS). Specificity was observed in motor, place, body, and most face localizers. Critically, on an individual-level, the specificity was significantly higher for the rsROIs than for the search spaces for all regions other than the OFA. Finally, we compare the specificity of these parcels to other similar parcels from previously released parcellations defined using functional connectivity.