Quantification of Tics in Tourette Syndrome

项目总结/摘要 抽动秽语综合征（TS）是一种慢性的、儿童期发作的神经发育障碍，影响1-3%的人 并且与不利的功能影响相关联。TS的特点是抽搐，这是不自主的，重复的， 动作和发声。TS临床护理的当前挑战是缺乏客观、定量、 测量抽搐的可扩展工具，用于诊断和症状严重性监测。整体 拟议研究的目的是在一个大的、多样化的社区样本中使用基于视频的方法， TIC的定量和自动化表型分析。本研究建立在以前的工作，包括：1）基于视频 观察方法与训练有素的人类评分员量化抽搐的研究目的，2）计算机视觉和 用于运动分析和医疗诊断辅助的机器学习技术，以及3）初步数据 这表明监督学习方法可以用于以高精度自动检测眼睛抽搐。在 目的1，将使用远程和远程技术收集N = 1，000名抽搐患者的视频和临床数据， 基于互联网的方法。将使用深度表型分析方法定量描述表型 可观察到运动和声音抽搐的频谱、经验性地导出的抽搐严重性基准以及识别患者 分组。在目标2中，我们的计算机视觉团队将对目标1数据应用监督机器学习方法 来创建一个能够检测最常见抽搐的算法。目标3将前瞻性地测试目标2 在一项单独的临床试验中完成TS治疗的N = 60例患者中使用算法，以确定算法的 变化敏感性和收敛效度与当前金标准抽动严重程度测量。这个项目 将使我们能够，第一次，在一个大的社区样本中量化可观察到的抽搐的频谱， 对诊断决策和患者教育具有直接临床相关性的知识。 目标2将产生一种计算机算法，能够自动量化最常见的运动抽搐， 关键的下一步是开发准确的，临床有效的，可扩展的评估抽搐筛查， 临床试验中的诊断、治疗决策和症状量化。

PROJECT SUMMARY/ABSTRACT Tourette Syndrome (TS) is a chronic, childhood-onset neurodevelopmental disorder that affects 1-3% of people and is associated with adverse functional impacts. TS is characterized by tics, which are involuntary, repetitive movements and vocalizations. A current challenge in clinical care for TS is the lack of objective, quantitative, scalable tools to measure tics for the purposes of diagnosis and symptom severity monitoring. The overall objective of the proposed study is to use video-based methods in a large, diverse community sample to inform quantitative and automated phenotyping of tics. This study builds on prior work, including: 1) video-based observational methods with trained human raters to quantify tics for research purposes, 2) computer vision and machine learning techniques for movement analysis and medical diagnostic aids, and 3) preliminary data indicating supervised learning methods can be used to automate detection of eye tics with high accuracy. In Aim 1, videos and clinical data from N = 1,000 individuals with tics will be collected using remote and internet-based methods. A deep phenotyping approach will be used to quantitatively describe the phenotypic spectrum of observable motor and vocal tics, empirically derive tic severity benchmarks, and identify patient subgroups. In Aim 2, our computer vision team will apply supervised machine learning methods to Aim 1 data to create an algorithm capable of detecting the most common tics. Aim 3 will prospectively test the Aim 2 algorithm in N = 60 patients who completed TS treatment in a separate clinical trial to establish the algorithm’s sensitivity to change and convergent validity with current gold-standard tic severity measurement. This project will enable us, for the first time, to quantify the spectrum of observable tics in a large community sample, knowledge that will have immediate clinical relevance for diagnostic decision making and patient education. Aim 2 will yield a computer algorithm capable of autonomously quantifying the most common motor tics, a critical next step toward developing accurate, clinically valid, and scalable assessments for tic screening, diagnosis, treatment decision making, and symptom quantification in clinical trials.