包含多类别信息的视觉物体的神经表征的功能磁共振研究

PSYCH OpenIR > 健康与遗传心理学研究室

	包含多类别信息的视觉物体的神经表征的功能磁共振研究
其他题名	Neural Representations of Visual Objects Containing Many Category Information: fMRI Functional Investigations
	张自富
	2013-06
摘要	对视觉物体神经表征的大量研究已经确定了大脑中存在若干表征类别特异性信息的神经系统。例如，面孔识别是由几个脑区完成的，包括梭状回面孔区（FFA）、枕叶面孔区（OFA）和颞上沟（STS）等；而场景信息主要在腹侧视觉系统的内侧部被表征，尤其是海马旁回位置区（PPA）。然而，在现实中，我们很少看见只包含单一类别信息的物体。面孔一般总是在某个背景中出现，而场景中又经常包含面孔。换而言之，在日常生活中，我们的视觉系统通常接受包含多类别信息的视觉物体刺激。而我们对于包含多类别信息的自然视觉物体的神经表征却所知甚少。比如，表征这些自然物体的神经系统是否和实验研究中发现的表征单一类别物体的神经系统相同？特别是，FFA和PPA在自然多类别物体的视觉加工中的作用是否和它们对单一面孔、单一场景视觉加工的作用相同？为了回答上述问题，我们采用功能磁共振成像考察大脑对四种类型的视觉刺激的神经活动：1) 存在于场景中的人物雕塑（以下简称雕塑）；2）巨型石刻（本身就是山景）（以下简称石刻）；3）单独的面孔；4）单独的山景。研究包括三个部分：第一部分采用广义线性模型（GLM）分析技术考察全脑对这些四类刺激的反应，我们要回答的问题是，雕塑和石刻激活的脑区是否和单独的面孔或单独的山激活的脑区相同？第二部分通过集中考察FFA和PPA这两个最具有代表性的面孔和场景类别选择区对四种视觉刺激的反应特征，区分它们在雕塑和石刻识别中的不同作用。我们首先采用GLM分析手段比较这两个脑区对雕塑和石刻相对于对面孔和山的激活幅度，然后采用表征相似性分析（RSA）和聚类分析测量多类别刺激（雕塑和石刻）和单一类别刺激（面孔和山景）在这两个脑区的多体素表征模式相似性程度。第三部分我们扩大研究视野，系统地刻画大脑对多类别自然视觉物体的层级表征，从初级视觉皮层、外侧枕叶皮层（LOC）、到OFA、再到梭状回中部的FFA和PPA。研究发现（1）雕塑和石刻都分别激活了腹侧视觉通络上的多个脑区，这些区域与单一面孔和单一山景激活的区域高度重叠，并且在梭状回中部存在一种一般的趋势，即大物体主要激活内侧部、而小物体主要激活外侧部，这种趋势与所包含的类别信息无关；（2）FFA对面孔、雕塑和石刻的反应幅度几乎相同，而PPA对山景、雕塑和石刻的反应幅度也几乎没有差异。但有趣的是， FFA和PPA 对这四种刺激的多体素反应模式明显不同；（3）GLM和RSA结果发现，从早期视觉皮层到梭状回中部，对多类别自然视觉物体选择性和模式鉴别力均逐渐增加。这些结果表明：（1）位于腹侧枕-颞皮层的各物体类别选择区协同工作来以加工包含多类别信息的自然物体，而各功能区加工多类别信息的方式似乎与加工它们偏好的单一类别信息没有本质性不同；（2）多体素的激活模式携带自然物体的多类别信息，并且从早期视觉皮层到梭状回中部物体选择区，信息的特异性水平越来越高。本研究还有许多有待于进一步解决的问题。比如，腹侧视觉系统的组织原则是“什么”？它为什么按照这种/些原则表征物体？物体大小和物体生命性之间的关系是什么？在今后的工作中，我们将采用更高高分辨率fMRI技术和更先进的分析手段研究这些遗留的问题。
其他摘要	Numerous laboratory studies have been conducted to investigate neural representations of visual objects and have identified a number of neural systems that represent category specific information. For example, face recognition is shown to be supported by several brain regions, including the fusiform face area (FFA), the occipital face area (OFA), and the superior temporal sulcus (STS). Similarly, scene information is represented in the medial part of the ventral visual system, particularly in the parahippocampal place area (PPA). However, in reality, objects are rarely viewed in isolation. Faces are usually embedded in backgrounds and scenes often contain faces. In other words, in daily life our visual system is typically presented with multi-category information. Unfortunately, little is known about the neural representations of natural visual objects containing multi-category information. Are such natural objects represented by the same neural systems as identified in the laboratory studies using isolated objects? More specifically, do the FFA and PPA play the same roles in viewing the natural multi-category visual objects as compared to viewing the isolated faces and scenes? To address these questions, we used functional magnetic resonance imaging (fMRI) to investigate neural activations to four types of visual stimuli: 1) human statues embedded in the context of scenes, 2) huge sculptures which were hills or mountains themselves, 3) isolated faces, and 4) mountain scenes. Our study consisted of three parts. In part one, we examined whole brain responses to these stimuli, to ask whether the statues and sculptures activated the same brain regions as the faces and mountains, using the General linear model (GLM) analysis. In part two, we focused on response properties of the FFA and PPA, the two most representative face and scene category selective regions, to delineate their functional role in the recognition of the statues and sculptures. We first used GLM analysis to compare the magnitude of activation to statues and sculptures with those of activations to faces and mountains. We next used representational similarity analysis (RSA) and cluster analysis to measure the degree of pattern similarity between multi-category stimuli (the statues and sculptures) and isolated stimuli (faces and mountains). In part three, we expanded our view and characterized the hierarchical representations of the natural visual objects, from the early visual cortex, lateral occipital cortex (LOC), OFA, to the FFA and PPA. We found: (1) the statues and sculptures activated several regions in the ventral visual pathway, largely overlapping with regions activated by faces and mountains; and there is a general trend of stronger activation towards the medial site for large size objects regardless of the category information contained; (2) FFA had approximately the same response magnitudes to faces, statues, and sculptures, and PPA had similar response magnitudes to mountains, statues, and sculptures; interestingly, the four types of stimuli had clear different multivoxel response patterns in the FFA and PPA; (3) both GLM and RSA revealed hierarchical features of representations of the natural visual objects: the selectivity and pattern discrimination gradually increased from the early visual cortex to the mid-fusiform cortex. These findings suggest: (1) the object category-selective areas in the ventral occipitotemporal cortex work together to process natural multicategory objects, with each area behaving not fundamentally different from processing their preferred objects in isolation;(2) the patterns of activation across multiple voxels contain distinct information about the natural multi-category objects, with increasing level of information specificity going from early visual cortex to the mid-fusiform object selective areas. However, many issues remain to be further addressed, such as the “what” and “why” of the principles of organization in the ventral visual system and the interaction between object size and object animacy, ideally with higher resolution fMRI and advanced analysis methods.
学科领域	认知神经科学
关键词	自然视觉物体神经表征梭状回面孔区旁海马回位置区功能磁共振
学位类型	博士
语种	中文
学位专业	心理学
学位授予单位	中国科学院研究生院
学位授予地点	北京
文献类型	学位论文
条目标识符	http://ir.psych.ac.cn/handle/311026/19776
专题	健康与遗传心理学研究室
作者单位	中科院心理研究所
推荐引用方式 GB/T 7714	张自富. 包含多类别信息的视觉物体的神经表征的功能磁共振研究[D]. 北京. 中国科学院研究生院,2013.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
张自富－博士学位论文.pdf（2070KB）	学位论文		限制开放	CC BY-NC-SA	请求全文