基于功能磁共振成像的立体图像分辨

Discrimination of Stereo Images by the Human Brain Based on fMRI

视觉是人与外界互动获取信息的主要手段,而双目视差信息是人脑估计外界环境深度结构的重要视觉线索之一.因此,研究人脑处理双目视差的神经机制对了解人类的视觉系统意义重大.功能磁共振成像(f MRI)技术为双目视差研究提供了有效手段.目前在f MRI研究中,虽然已经有很多研究利用f MRI技术深入探究了人脑处理双目视差信息的神经机制,但是利用该技术采集的人脑信号如何分辨包含双目视差信息的立体图像依然有待研究.针对这一问题,设计了一种基于f MRI的实验,该实验选用随机点图生成人造立体视图像作为实验刺激;相较于自然立体图像,该种立体图像可以更加方便地提取出立体图像中包含的图像特征.结合实验特性提出了一种基于lasso回归算法的体素编码模型,该模型利用了视觉感受野的稀疏特性,可以较好地借助立体图像中的二维特征并对fMRI数据进行编码分析和解码分析.其编码分析结果表明利用体素编码模型可以较好预测人脑接收立体图像的脑信号的体素广泛分布在人脑的各个视觉区中,并且大部分体素分布在初级视区V1、V2d和V3d中.解码分析结果表明,初级视觉区V1可以利用立体图像中的二维特征实现立体图像的识别,并且背侧视觉区V3d、V7和h MT+/V5可以与V1协同工作进行立体图像的识别.

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 2 D 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,V2 d and V3 d.The decoding analysis results indicated that primary visual area V1 had higher discriminative power to stereo image by using 2 D features of stereo images.Moreover,the dorsal visual areas V3 d,V7 and hMT+/V5 could work with V1 to discriminate stereo images.