Item

Abstract

Recent studies have demonstrated that multivariate machine learning algorithms applied to human functional MRI data can decode information segregated in cortical columns, despite the voxel size being large relative to the width of columns. The mechanism by which low spatial resolution imaging decodes information represented in a fine-scale organization is not clear.

To investigate mechanisms underlying decoding signals we employed high-resolution gradient-echo BOLD functional MRI of visual area V1. We show that in addition to the fine-scale ocular dominance columns, coarse-scale structures extending over several millimeters also convey discriminative power for decoding the stimulated eye. Discriminative power is conveyed by both macroscopic blood vessels and gray matter regions. We hypothesize that gray-matter regions which drain into specific vessels may preferentially contain ocular-dominance columns biased towards one eye; the bias of a specific region thereby causing a functionally selective ocular-dominance response in the associated vessel. Our findings indicate that coarse-scale structures and macroscopic blood vessels contribute to decoding of the stimulated eye based on low-resolution multivariate data.