Li Zhen, Shigemasu Hiroaki
Graduate School of Engineering, Kochi University of Technology, Kochi, Japan.
School of Information, Kochi University of Technology, Kochi, Japan.
Front Hum Neurosci. 2019 Aug 20;13:283. doi: 10.3389/fnhum.2019.00283. eCollection 2019.
The brain's ability to extract three-dimensional (3D) shape and orientation information from viewed objects is vital in daily life. Stereoscopic 3D surface perception relies on binocular disparity. Neurons selective to binocular disparity are widely distributed among visual areas, but the manner in these areas are involved in stereoscopic 3D surface representation is unclear. To address this, participants were instructed to observe random dot stereograms (RDS) depicting convex and concave curved surfaces and the blood oxygenation level-dependent (BOLD) signal of visual cortices was recorded. Two surface types were: (i) horizontally positioned surfaces defined by shear disparity; and (ii) vertically positioned surfaces defined by compression disparity. The surfaces were presented at different depth positions per trial. Functional magnetic resonance imaging (fMRI) data were classified from early visual areas to higher visual areas. We determined whether cortical areas were selective to shape and orientation by assessing same-type stimuli classification accuracies based on multi-voxel activity patterns per area. To identify whether some areas were related to a more generalized sign of curvature or orientation representation, transfer classification was used by training classifiers on one dataset type and testing classifiers on another type. Same-type stimuli classification results showed that most selected visual areas were selective to shape and all were selective to the orientation of disparity-defined 3D surfaces. Transfer classification results showed that in the dorsal visual area V3A, classification accuracies for the discriminate sign of surface curvature were higher than the baseline of statistical significance for all types of classifications, demonstrating that V3A is related to generalized shape representation. Classification accuracies for discriminating horizontal-vertical surfaces in higher dorsal areas V3A and V7 and ventral area lateral occipital complex (LOC) as well as in some areas of intraparietal sulcus (IPS) were higher than the baseline of statistical significance, indicating their relation to the generalized representation of 3D surface orientation.
大脑从所观察的物体中提取三维(3D)形状和方向信息的能力在日常生活中至关重要。立体3D表面感知依赖于双眼视差。对双眼视差具有选择性的神经元广泛分布于视觉区域,但这些区域参与立体3D表面表征的方式尚不清楚。为了解决这个问题,研究人员指示参与者观察描绘凸面和凹面曲面的随机点立体图(RDS),并记录视觉皮层的血氧水平依赖(BOLD)信号。两种表面类型分别为:(i)由剪切视差定义的水平放置表面;以及(ii)由压缩视差定义的垂直放置表面。每次试验中,这些表面呈现于不同的深度位置。功能磁共振成像(fMRI)数据从早期视觉区域到高级视觉区域进行分类。我们通过基于每个区域的多体素活动模式评估同类型刺激分类准确率,来确定皮层区域是否对形状和方向具有选择性。为了确定某些区域是否与更广义的曲率或方向表征信号相关,采用了转移分类,即使用一种数据集类型训练分类器,并在另一种类型上测试分类器。同类型刺激分类结果表明,大多数选定的视觉区域对形状具有选择性,并且所有区域对视差定义的3D表面的方向都具有选择性。转移分类结果表明,在背侧视觉区域V3A中,所有类型分类的表面曲率辨别信号的分类准确率均高于统计学意义基线,这表明V3A与广义形状表征相关。在较高的背侧区域V3A和V7、腹侧区域外侧枕叶复合体(LOC)以及顶内沟(IPS)的某些区域中,辨别水平 - 垂直表面的分类准确率高于统计学意义基线,表明它们与3D表面方向的广义表征有关。