情境驱动的多动症可解释量化评估方法研究

Attention Deficit Hyperactivity Disorder (ADHD) is one of the most common mental disorders in childhood. Because the etiology of ADHD is complex and the symptoms are not specific, there is still a lack of feasible quantitative diagnostic methods. Pursuing objective and non-invasive detection methods and standards is of great practical significance to prevent the development of the disease. Aiming at the objective quantitative diagnosis of ADHD, this project investigates the context-driven explainable and quantitative assessment method. Based on the international standard classification of mental disorders DSM-5 and doctors'rich clinical experience, 1) this project designs and implements the scale-driven detection scenario design and wearable multimodal data perception, and constructs the ADHD activity sample dataset. 2) Through the specific feature extraction and feature sharing of context fusion, the project realizes the fine-grained assessment model for the ADHD subtypes. 3) At the same time, according to the interpretability of symptoms, a quantitative correlation model for ADHD activity and cognition is studied.And finally an early assessment system for school-age children is constructed. This project combines the Institute of Computing Technology, Chinese Academy of Sciences and Beijing Anding Hospital, affiliated to Capital Medical University, to give full play to the advantages of both disciplines, and improves the clinical diagnosis rate of ADHD. It also provides an important practical reference for the application of artificial intelligence (AI) in the field of perception and cognition.

注意缺陷多动障碍（ADHD）是儿童时期最为常见的精神障碍类疾病之一，由于ADHD的病因复杂，症状不具有显著的特异性，目前仍缺乏可行的量化诊断方法。因此，追求客观的、无损伤的检测手段和检测标准对于阻止病情的发展有重要的现实意义。本项目针对ADHD的客观量化诊断，研究情境驱动的多动症可解释量化评估方法。基于国际精神障碍通用量表DSM-5诊断标准和医生丰富临床经验，1）设计实现量表驱动的多种检测场景及可穿戴多模数据感知，并构建ADHD行为样本数据集；2）通过情境融合的特异性特征提取和共享实现面向ADHD亚型诊断的细粒度评估模型构建；3）针对症状的可解释性，研究面向ADHD行为与认知的量化关联模型。最终构建面向学龄期儿童的ADHD早期评估系统。本项目将联合中科院计算所和首都医科大学附属北京安定医院，发挥双方学科优势，提高ADHD的临床诊断率，并为人工智能在感知和认知领域的应用提供重要的实践参考。

注意缺陷多动障碍(ADHD)是影响儿童最常见的精神障碍之一。针对ADHD病因复杂，症状不具特异性，缺乏可行的定量诊断方法的问题，为寻求客观、无创的检测方法和标准，预防该病的发展，本项目开展了情境驱动的多动症可解释量化评估方法研究。研究实现了量表驱动的多模数据协同感知、面向ADHD亚型诊断的细粒度评估以及面向ADHD行为与认知的量化关联方法。相关成果在IEEE International Conference on Ubiquitous Intelligence and Computing（IEEE IMWUT）、CHI Conference on Human Factors in Computing Systems、ACM Transactions on Intelligent Systems and Technology（ACM TIST）等国内外知名期刊会议上发表和接收论文11篇，申请国家发明专利11项、软件著作权2项目，发布白皮书1项。共参加国际学术会议5人次。合作培养4名博士研究生和5名硕士研究生。