Cortical organization of binocular disparity in human V3A

Abstract

Our current knowledge of the columnar architecture of the cortex relies mainly on animal models, while much less is known about cortical organization in the human brain. The major barrier in understanding the fine-scale human brain computations comes from limitations in data acquisition. Standard fMRI resolution at 3T allows us to study large neural populations but not the microstructure. Here we use high-field (7T) functional magnetic resonance imaging (fMRI) to test for columnar organization for binocular disparity in human visual area V3A, an area previously associated with depth perception. We presented participants (N=5) with disparity-defined wedges ranging from fine to coarse disparity magnitudes (3 to 36 arcmin, crossed and uncrossed). During stimulus presentation, participants performed a demanding Vernier detection task at the central fixation point. We measured BOLD responses from dorsomedial visual cortex using 0.96x0.96x1 mm voxels and a 3D gradient echo sequence. We estimated disparity selectivity using general linear modeling, and then characterized the organization of voxel preferences across the cortical surface. First, we found that similar disparity preferences were clustered together to form structured maps that were reproducible across two different scan sessions. Then, we modeled individual voxel responses using Gabor-based filters inspired by electrophysiological recordings. We found that individual voxels were more frequently described as tuned to particular disparity values for fine disparities; however, when coarser disparities were tested, more voxels responded in a categorical (near vs. far) manner. These preferences for different types of voxel response (tuned vs. categorical) were clustered together on the cortical surface. Thus we provide evidence that human V3A is systematically organized to represent binocular disparity information, highlighting the importance of this area for stereopsis in the human brain.

Meeting abstract presented at VSS 2014