基于脑电分析与稀疏编码的ADHD听觉选择性注意的认知神经机制研究

Attention deficit/hyperactivity disorder (ADHD）is one of the prevalent psychiatry disease, which has great impacts on patients' daily living and social security. Auditory selective attention (ASA) is not only the cognitive processes of learning, judgment and decision, but also the research basis for human-machine interaction. However, the quantity clinical criteria for auditory neural deficient in ADHD are not sufficient, and the neural mechanisms of ASA in ADHD are not fully understood. The purpose of this project is to investigate the difference of neural and behavioral signals between ADHD patients and healthy control subjects when they implement ASA task, then explore the neural mechanisms of spatial and temporal auditory attention in ADHD, and provide solid clinical diagnostic criteria for auditory attention deficiency of ADHD. Behavioral and electroencephalogram (EEG) experiments are carried out in our research. Firstly, EEG waves, EEG power, characteristics of brain region, response time, and behavioral correction rate will be provided as the clinical criteria for ADHD diagnose. Then, the ADHDs' auditory neural phase locking to temporal envelope and fine structure of Mandarin will be analyzed, in order to give instruction on language rehabilitation for ADHD. Finally, by means of matching pursuit, particle filter and other approaches, the study also focuses on two auditory neural sparse coding models of ADHD in the ASA task, including time-frequency coding and temporal dynamic coding. These two spare coding models are helpful for investigating the auditory neural mechanisms and deficiencies of ADHD patients.

注意缺陷多动障碍（ADHD）是常见的精神类高发疾病之一，严重影响患者的日常生活，甚至影响社会安全。听力选择性注意（ASA）不仅决定了人的学习、判断、决策等认知能力，也是人机交互研究的基础。然而，ADHD的听觉神经缺陷仍没有定量诊断标准，且患者的ASA的神经机制仍未知。本项目拟通过脑电和行为学实验，分析ADHD 患者与健康对照组执行ASA 任务时的脑电神经信号和行为，探索ADHD患者的空间和时间听觉选择性注意力的神经机制及临床定量诊断方法。首先，提供ADHD的脑电波形、脑电响应能量、脑区特征、响应时间、行为准确率等临床诊断指标。其次，分析患者对中文语音的时域包络、时间细节结构的听觉神经元锁相能力，目的是为言语康复训练提供依据。最后，用匹配追踪、粒子滤波等方法，建立ADHD患者的ASA听觉稀疏神经的时频域和时域动态编码的两个模型，更全面地揭示患者的听觉神经通路的工作机制和可能存在的缺陷。

本申请项目通过听觉脑电实验、行为学实验和脑部磁共振图像，完成了一系列分析ADHD患者神经信号和行为数据的算法和数学模型，探索了ADHD患者的注意力的神经机制，为临床提供了量化诊断ADHD的可靠方法。项目完成了以下七个内容：听觉选择性注意实验及可塑性神经机制研究、中文听觉选择性注意实验平台构建及测试、噪声环境中的听觉频率跟随响应研究、ADHD患儿CPT脑电事件相关电位研究、ADHD患儿脑电信号双谱研究、基于人工智能算法的ADHD患儿脑部图像诊断、基于机器学习算法的ADHD患儿脑部EEG病类判别。项目招募ADHD受试者和正常儿童共计完成了16个实验。项目提出了12种新方法，包括：4种分析FFR的幅度与相位的算法；3种分析ERP的双谱特征与相位的方法；5种MRI的分割与分类相关的新算法。完成了项目全部研究内容。完成了相关论文及专利21项，获批主持相关科研项目12项，产学研教学改革项目4项，培养学生获奖等11项。该项目的结论已应用于ADHD临床诊断，提出的方法已应用于产学研系统。