Discrimination of Stereo Images by the Human Brain Based on fMRI

Vision is a main pathway for human understanding of the physical world. Binocular disparity is an important visual cue in the human brain's estimation of depth structure in the environment. Therefore, exploring the neural mechanism of disparity processing is essential for understanding the human visual system. Functional magnetic resonance imaging(fMRI)technology provides an effective method for the study of binocular disparity in humans. Although some fMRI studies have explored the neural mechanism of binocular disparity processing, it has remained unclear as to how to discriminate stereo images from the observed fMRI data. In this study, an fMRI experiment was designed and the artificial stereo images by random dot stereograms were used as experimental stimuli. Compared with the natural stereo image, the image features in the artificial stereo image can be extracted more easily. Considering the characteristics of the experiment designed, a voxel-wise encoding model based on lasso was proposed. The model with the sparsity of the visual receptive field can utilize the 2D visual features of stereo images to perform encoding and decoding analyses from fMRI data. The encoding analysis results indicated that the voxels were distributed in various visual areas, which could accurately predict the human brain signals from stereo images by voxel-wise encoding models. Most of these voxels were distributed in the primary visual areas V1, V2d and V3d. The decoding analysis results indicated that primary visual area V1 had higher discriminative power to stereo image by using 2D features of stereo images. Moreover, the dorsal visual areas V3d, V7 and hMT+/V5 could work with V1 to discriminate stereo images. © 2019, Editorial Board of Journal of Tianjin University(Science and Technology). All right reserved.