Abstract
Recent studies of primate brain showed that almost all areas of the visual cortex contain neurons that respond to binocular disparity. However, the functional differences among these areas are largely unknown. From binocular disparity, we perceive three-dimensional spatial layouts of multiple objects and three-dimensional shape of a single object. We hypothesize that these two types of the depth perception from binocular disparity are carried out by different cortical areas. This classification corresponds to the distinction between a dorsal pathway that concerns with spatial relationships of objects and a ventral pathway that concerns with object shape. We used the decoding technique (Kamitani & Tong, 2005) and compared the prediction accuracy of these two types of depth perception between cortical activities of dorsal and ventral areas. fMRI signals (1.5 T, 3 × 3 × 3 mm voxels) were measured while subjects viewed random dot stereograms. For the spatial depth perception among multiple objects, we showed three independent surfaces with different depth planes. The center surface was closer or further from the other two surfaces which were positioned on the same depth plane. For the three-dimensional shape perception, we showed a single convex or concave curved surface. A linear decoder (support vector machine) was trained to classify voxel intensity patterns induced by far/near position or convex/concave shape. Then the decoder was evaluated with independent test data. In results, area hMT+ showed better performance for classification of depth order than that of three-dimensional shape. In contrast, area LOC showed better performance for three-dimensional shape than depth order. These results suggest that the dorsal pathway area hMT+ concerns with the perception of three-dimensional layouts of multiple objects and the ventral pathway area LOC concerns with the perception of three-dimensional shape from binocular disparity.
This study was supported by the global COE Program ‘Frontiers of Intelligent Sensing’ from MEXT Japan and Nissan Science Foundation.