基于神经电信号的信息反馈计算模型及应用

Neuropsychical diseases pose a serious threat to human health nowadays. Neurofeedback, which aims to self-control brain activity based on operant conditioning, has been shown to be effective as adjunctive therapy for these diseases. However, little is known about computational models and neurophysiological mechanisms of neurofeedback due to limitations of current brain monitoring techniques. This project proposes to comprehensively study theory and methodology of Electroencephalography (EEG)-based neurofeedback through three subtopics. The first subtopic will focus on developing computational models for EEG-based neurofeedback. The proposed training paradigm is to learn to self-regulate brain state, which can be decoded with multi-channel EEG signals in real time. Specifically, this subtopic aims to develop methodologies to achieve fast and reliable control of brain states and assess the efficacy of the brain-state based neurofeedback training on regulating cognition and behavior. The second subtopic will develop a general neurofeedback platform technology for method validation and real applications. This project will apply the target coding and decoding techniques in current brain-computer interface (BCI) systems to develop new data analysis methods and design multi-layered and multi-targeted feedback paradigms. Robustness and speed of the decoding method will be improved by modeling co-adaptive learning between the brain and the system. In addition, this subtopic will investigate quantitative approaches to combine neural and behavioral data to trace and access cognitive and cortical effects in neurofeedback training. The third subtopic will aim to apply the proposed new methodologies to the adjunctive treatment of depression in clinically depressed patients. The goal of this subtopic is to establish an optimal training paradigm for the treatment of depression and assess the effects on cognitive enhancement after training.

神经精神类疾病是一种严重威胁人类健康的疾病。神经反馈可以通过自主调节脑活动达到辅助治疗的目的。受脑监测技术的限制，人们对神经反馈的计算模型和神经生理机制知之甚少。本项目通过以下三个子课题来深入研究基于脑电信号的神经反馈的理论和方法。（1）在计算模型方面，本项目针对多电极的脑电信号进行实时脑状态解码；通过可视化方法呈现给用户，建立快速有效地调控不同皮层神经活动的策略；评估神经反馈调节、改善认知行为的有效性。（2）在神经反馈实验平台方面，本项目借鉴脑机接口的信息编码和解码的原理，建立神经反馈的分析方法实现多层次、多靶点的反馈模式。对大脑系统与解码算法的互适应关系进行建模，使解码算法更加稳健快速。同时结合其它行为学数据定量跟踪和评价反馈训练带来的脑功能的改变与提升。（3）在神经反馈的应用方面，本项目建立适用于抑郁病患者认知功能增强的神经反馈训练系统，评估反馈训练对抑郁病患者认知功能增强的效果。

神经反馈对神经精神类疾病的治疗和健康人群的认知提升具有重要意义。然而，当前神经反馈存在神经机理不明、计算分析模型缺失、靶点和反馈策略缺乏定量评估的研究难点，造成了神经反馈系统的性能瓶颈和应用困境。从“脑”到“机”和从“机”到“脑”，作为双向神经信息编码和神经调控的一体两翼，脑-机接口技术已成为进一步提升神经反馈性能和效果的重要发展内驱力和未来趋势，本项目重点关注基于脑-机接口的神经反馈系统，建立理论计算模型，实现高性能的脑-机交互，完成可实用的认知功能训练系统并开展临床测试。.研究内容分为基于EEG神经反馈的理论计算模型研究、基于EEG神经反馈的实验平台研究、面向认知和运动增强的神经反馈训练系统研究三部分。.（1）基于EEG神经反馈的理论计算模型：项目研究了稳态视觉诱发电位（SSVEP）动态模型、稳态刺激噪声和疲劳对SSVEP的影响、多导脑电数据中诱发电位检测的空时均衡方法。计算模型可帮助设计后续反馈系统，提高基于EEG的神经反馈实验平台的性能。.（2）基于EEG神经反馈的实验平台研究：项目研发了基于非侵入式脑-机接口的高速率字符输入系统、结合脑-机接口和计算机视觉的机械臂控制系统、一种基于SSVEP的人机协同控制系统、在线SSVEP-BCI异步系统。在系统通信速率上实现了世界最快的BCI打字系统，且进一步在机械控制、人机协同以及异步实际应用中进行了扩展优化。.（3）面向认知和运动增强的神经反馈训练系统研究：项目研究了基于神经反馈训练提升儿童注意及SSVEP脑-机接口的实用效果、阳极tDCS及tACS对SSVEP的调制作用、神经反馈训练对抑郁症患者的工作记忆注意和执行功能的增强效果。发现了基于提高alpha振荡幅度、经颅电刺激可提高SSVEP脑电响应，以及神经反馈辅助治疗可以提高抑郁状态患者的运动认知和运动学习能力。.项目执行期间，共发表论文55篇，其中SCI收录34篇，国际会议15篇，中文核心期刊6篇。标注本自然科学基金项目资助论文55篇，标注率100%。申请发明专利21项（授权13项），软件著作权1项。.