Understanding the Conceptual Priority Map Guiding Naturalistic Visual Attention for Autistic Individuals

项目摘要 视觉注意差异是自闭症谱系疾病(ASC)的一个很有前途的诊断标志。然而， 尽管越来越多的证据表明视觉注意在群体水平上存在差异，特别是在视觉注意定向方面 关于自闭症患者和非自闭症患者之间的社会相关信息(即“社会凝视”)，来源 自闭症患者的凝视差异仍不清楚。关于社会凝视差异的著名理论主要集中在 特定类别的视觉刺激，即：人脸。这些理论没有回答的是，是否会减少 事实上，社会关注最好的解释是对特定刺激类别的非典型关注，或者它是否反映了 从根本上减少了对重要社交信息的分布式(人脸和非人脸)来源的关注 复杂的环境。换句话说，社交信息可能不限于单一的视觉类别，而且它 在性质上可能根本不是绝对的。然而，由于专注于物体类别，眼球追踪分析失败了 要捕捉真实世界环境的丰富性和复杂性，其中视觉注意力支持 个人的行为。为了利用视觉注意作为一种临床上可操作的工具，关键的知识缺口 必须解决的问题是：自闭症的社会凝视差异是由视觉类别水平的信息驱动的吗， 或者取而代之的是超越视觉领域的更高层次的概念信息？ 这个项目的目标是检查分类和概念层面的影响 关于个体和自闭症群体在视觉注意力上的差异的信息。中心假设是视觉上的 自闭症患者的注意力差异源于概念水平而不是类别水平的心理差异 正在处理。为了验证这一假设，我开发了一种使用计算机视觉工具的新方法 (计算神经网络；CNN)和自然语言处理(NLP)来单独描述 独特的视觉注意力模式。首先，特定目标1a将测试凝视模式是否反映高维度 个别参与者独特的概念优先事项(N=62名非自闭症成年人)。具体目标1b将 测试概念优先级是否可靠地指导自闭症患者(N=28)的凝视，并可用于分类 按诊断状态划分的个人(自闭症与非自闭症)。具体目标二，博士后研究方向， 我将把我的论文研究的重点，关于驱动视觉注意的概念优先级，扩展到概念 视觉领域之外的优先事项，例如语言。这些目标已经作为结构化的 培训计划旨在促进向博士后职位和独立研究职业的过渡。这 培训计划强调多元统计分析、实验设计和科学的技能发展 沟通。这项培训计划由卡罗琳·罗伯逊博士赞助，她在自闭症方面的专长是视觉 处理和新的实验技术(例如，虚拟现实)理想地由技术和 达特茅斯大学心理和脑科学系的计算能力。

Project Summary Visual attention differences are a promising diagnostic marker for autism spectrum conditions (ASC). Yet, despite mounting evidence for group-level differences in visual attention, particularly for visual attention directed toward socially relevant information (i.e., “social gaze”) between autistic and non-autistic individuals, the source of gaze differences in autism remains unclear. Prominent theories of social gaze differences focus heavily on a particular category of visual stimuli, namely: faces. What these theories leave unanswered is whether reduced social attention is, in fact, best explained by atypical attention to a specific stimulus class or whether it reflects an underlying reduction in attention to distributed (face and non-face) sources of important social information in complex environments. In other words, social information may not be limited to a single visual category, and it may not be categorical in nature at all. Yet, by focusing on object categories, eyetracking analyses have failed to capture the richness and complexity of real-world environments in which visual attention supports an individual’s behavior. In order to leverage visual attention as a clinically actionable tool, a critical knowledge gap must be addressed: are social gaze differences in autism driven by information at the level of visual categories, or instead, by higher-order conceptual information beyond the visual domain? The objective of this project is to examine the impact of both categorical and conceptual levels of information on individual and autistic group differences in visual attention. The central hypothesis is that visual attention differences in autism stem from conceptual-level, rather than categorical-level, differences in mental processing. To test this hypothesis, I have developed a novel approach that uses tools from computer vision (computational neural networks; CNNs) and natural language processing (NLP) to characterize individually unique patterns of visual attention. First, Specific Aim 1a will test whether gaze patterns reflect high-dimensional conceptual priorities that are unique to individual participants (N = 62 non-autistic adults). Specific Aim 1b will test whether conceptual priorities reliably guide autistic individuals’ (N = 28) gaze and can be used to classify individuals by diagnostic status (autistic vs. non-autistic). Specific Aim 2, the postdoctoral research direction, will extend the focus of my dissertation research, on conceptual priorities that drive visual attention, to conceptual priorities outside the visual domain, such as language. These aims have been articulated as part of a structured training plan designed to facilitate the transition to a postdoctoral position and independent research career. This training plan emphasizes skill development in multivariate statistical analysis, experimental design, and scientific communication. This training plan is sponsored by Dr. Caroline Robertson, whose expertise in autism, visual processing, and novel experimental techniques (e.g., virtual reality) is ideally complemented by the technical and computational strengths in the Psychological and Brain Sciences Department at Dartmouth.