Abstract
Psychophysical studies have demonstrated a special role for curved (as compared to straight) contours. Physiological recordings have demonstrated a significant percentage of cells tuned to curved edges in V4d of macaques. Recently, human fMRI data demonstrated a segregation of activity elicited by curved vs. rectilinear stimuli, which partially overlaps the organization of face vs. scene driven activity. Based on this evidences, we hypothesized that: 1) curvature processing would be anatomically segregated in macaque visual cortex, perhaps including V4; 2) curvature processing in macaques would be anatomically linked with face processing regions. Functional MRI was acquired in four fixating macaques (4.7T, MION, 1.5[sup]3[/sup] mm voxels), in response to round vs. rectilinear shapes of different types, matched to those used in human fMRI. Stimuli included: 1) images of rounded vs. rectilinear real world objects, 2) computer-generated arrays of 3D shapes (e.g. spheres, or pyramids). These shape-based stimuli included no faces or scenes. The known face-selective cortical areas were localized in independent runs. The results revealed discrete regions that were especially well activated by the curved shapes, in both hemispheres of all animals. One region (posterior curvature region, PCR) was located in the vicinity of the foveal representation of V4, on the anterior portion of the prelunate gyrus. A second region (middle curvature region, MCR) was located ~3.1 cm anterior to PCR within the lower bank of the STS. The third and more weakly activated region (anterior curvature region, ACR) was located near the tip of the anterior temporal lobe. The extent (but not the peaks) in MCR and ACR partially overlapped with known face-selective regions. In further experiments we used the banana filters (a variant of Gabor wavelets specifically tuned to curvature) to quantify stimulus image curvature. The mean banana filter values significantly correlated with fMRI responses across conditions in PCR, suggesting that PCR processes simple curvature (r=0.79)
Meeting abstract presented at VSS 2013