A machine learning computational approach for developing synchronized EEG and behavior biomarkers in young autistic children

摘要-项目3 杜克大学自闭症卓越中心的总体目标是使用翻译的数字健康和计算 解决对更有效的自闭症筛查工具、客观结果衡量标准、 以及基于大脑的生物标记物，可用于自闭症儿童的临床试验。尽管意义重大 了解自闭症的生物学基础的进展，临床试验仍然依赖于主观的临床 观察和照顾者报告测量。客观地，基于生物的生物标记物需要用于 可以分析异质性、评估目标参与和监测结果的临床试验。自闭症生物标志物 研究利用了脑电(EEG)和眼球跟踪测量，发现了不同之处 自闭症患者和神经典型个体在社会刺激的神经和注意加工方面的差异。然而，为了 迄今为止，大多数自闭症生物标记物的研究都使用了独立的实验范式和独立的 脑电和凝视的分析。技术和计算方面的进步，包括机器学习和计算机 视觉分析，现在允许同步测量和分析脑电和行为，包括眼睛- 跟踪，每个跟踪都提供不同的信息源，可以集成这些信息源以改进生物标记物 性能。项目3将使用一种创新的机器学习计算方法来开发多模式 结合脑电活动特征和儿童行为同步测量的生物标记物(例如，社交 注意)通过计算机视觉分析自动编码。我们将检验这样的假设，即多模式 与生物标记物相比，生物标记物将显示自闭症儿童和神经症儿童之间更强的辨别力 仅凭脑电就可以。将使用标准和新方法来结合同步行为(数字 表型)和脑电特征，重点是通过传统方法测量的神经连接 (相干性、相位滞后指数)和新的网络分析方法(区分性交叉谱因子分析) 由我们的团队开发。这种多模式方法将在3-6岁的自闭症儿童中进行评估， 智障(ID)，年龄和性别匹配的神经典型儿童，以及智商=70的自闭症儿童。 多峰生物标记物将与三种常用的脑电生物标记物进行比较。我们的目标是发展强大的、 可用于临床试验以评估儿童自闭症早期干预的脑基生物标志物 旨在改善结果和生活质量。

ABSTRACT – Project 3 The overall goal of the Duke Autism Center of Excellence is to use a translational digital health and computational approach to address the critical need for more effective autism screening tools, objective outcome measures, and brain-based biomarkers that can be used in clinical trials with young autistic children. Despite significant advances in understanding the biological basis of autism, clinical trials continue to rely on subjective clinical observation and caregiver report measures. Objective, biologically based biomarkers are needed for use in clinical trials that can parse heterogeneity, assess target engagement, and monitor outcomes. Autism biomarker studies have utilized electroencephalography (EEG) and eye-tracking measures, which have found differences between autistic and neurotypical individuals in neural and attentional processing of social stimuli. However, to date, the majority of autism biomarker studies have used independent experimental paradigms and separate analyses of EEG and gaze. Technical and computational advances, including machine learning and computer vision analysis, now allow for synchronized measurement and analysis of EEG and behavior, including eye- tracking, each of which provides distinct sources of information that can be integrated to improve biomarker performance. Project 3 will use an innovative machine learning computational method to develop a multimodal biomarker that combines features of EEG activity and synchronized measures of children’s behavior (e.g., social attention) automatically coded via computer vision analysis. We will test the hypothesis that a multimodal biomarker will show enhanced discrimination between autistic and neurotypical children compared to biomarkers based on EEG alone. Standard and novel methods will be used to combine synchronized behavior (digital phenotypes) and EEG features, with a focus on neural connectivity measured via traditional methods (coherence, phase-lag index) and new network analysis methods (discriminative cross-spectral factor analysis) developed by our team. This multimodal approach will be evaluated in 3–6-year-old autistic children without intellectual disability (ID), age- and sex-matched neurotypical children, and autistic children with ID (IQ <= 70). Multimodal biomarkers will be compared to three commonly used EEG biomarkers. Our goal is to develop robust, brain-based biomarkers that can be used in clinical trials to evaluate early interventions for young autistic children designed to improve outcomes and quality of life.