Poster Session C, Sunday, March 24, 5:00 – 7:00 pm, Pacific Concourse
fMRI pattern similarity analyses reveal working memory and perceptual coding at both regional and brain-network level
Maria Z. Gehred1, Joset A. Etzel1, Todd S. Braver1; 1Washington University in St. Louis
Pattern similarity analyses have been used to identify coding properties of different brain regions, but rarely to compare regional and network-level properties in higher cognitive domains. In the current study, we leverage the Human Connectome Project (HCP) dataset and an established cortical parcellation scheme (Gordon atlas) to examine the coding of working memory (WM) load (0-back, 2-back) versus perceptual category (Face, Place) in the N-back task. We compared effects observed at the community level (FrontoParietal, Visual) with a whole-brain analysis focused on cortical parcels in order to determine relative sensitivities. As expected, we observed a robust dissociation at the community level, with FrontoParietal strongly reflecting WM load-based coding, and Visual reflecting perceptual coding. Community-level analyses seemed most robust for perceptual coding, as most of the parcels within the Visual community (28 of 39) showed weaker coding than the community as a whole. Likewise, across the whole-brain, only 8 parcels, all located within the Visual community, showed stronger coding than the community. A different picture emerged when examining load-based coding, as most FrontoParietal parcels (14 of 24) showed equivalent or stronger coding than the FrontoParietal community as a whole. A number of cortical parcels in other communities related to working memory (Dorsal Attention, Cingulo-Opercular) also displayed stronger coding. Additionally, the strength of load-based coding in these parcels predicted better behavioral performance (d’). These results highlight the utility of pattern similarity analyses for examining not only perceptual coding, but also the coding of working memory load at both network and regional levels.
Topic Area: METHODS: Neuroimaging