非眼动型脑-机接口中隐性视觉空间注意的神经学模型与识别方法研究

Brain-computer interfaces (BCIs) have been demonstrated to be powerful in the application of neural rehabilitation. BCIs based on visual event-related potentials (ERPs) are famous for its high speed and large instruction set. However, the traditional gaze-dependent visual BCI, requiring users to adjust the vision focus by eye movements, is not a direct translation of brain intents. The advanced cognitive process only plays a minor part in the neural feedback training in the traditional approach, which is not good for neural rehabilitation. Therefore, the gaze-independent BCI based on the direct measurement of covert visuospatial attention (CVSA), a kind of advanced cognitive process, is desiderated. This project aims to develop a CVSA-based gaze-independent BCI using the hybrid features of steady-state visual evoked potentials (ssVEPs) and the N2pc which is an earliest endogenous component reflecting the visuospatial attention in the ventral stream. The combination of ssVEPs and N2pc could provide more abundant, stronger and higher signal-to-noise rate cognitive EEG features for mental states decoding. This project includes studying the fundamental laws of the ssVEP and N2pc hybrid features under the modulation of CVSA, exploring N2pc’s neural mechanism and constructing its evolution model, developing a new algorithm for fast decoding the endogenous attention state, and building a high-efficiency gaze-independent BCI with high speed and high accuracy. The new system could work more effective for neural rehabilitation and this project could also provide a new theory and method for BCI developments.

脑-机接口（BCI）在神经康复应用显示强大潜力。视觉响应诱发事件相关电位（ERP）型BCI有速度快、指令多等优点，但传统诱发范式需用户调整视线方向、用显性空间注意力选择目标指令，严重妨碍大脑控制意图直接表达、削弱高级认知反馈训练、制约应用效果，亟需研发直接检测大脑内源性隐性视觉空间注意力（CVSA）信息的非眼动依赖视觉响应型BCI新技术。项目采用与选择注意密切相关、能最早体现视觉注意、有鲜明偏侧性的短时内源性ERP成分N2pc和稳态视觉诱发电位（ssVEP）相结合的方式构建非眼动型BCI诱发范式以获得有用信息量大、信噪比高、相关特征强的认知脑电，研究隐性视觉空间注意调制N2pc与ssVEP混合脑电特征基本规律，探索内源性脑电成分N2pc形成神经机制并建立演进模型，发展快速解码大脑内源性注意力意图的新算法，为神经康复应用提供正确率高、速度快的非眼动新型BCI系统，为其发展提供新理论、新方法。

视觉型脑-机接口（BCI）有速度快、指令多等优点，但传统诱发范式需用户调整视线方向、用显性空间注意力选择目标指令，严重妨碍大脑控制意图直接表达。针对该问题，本项目研究了可用于BCI控制的大脑隐性视觉空间注意力（CVSA）新方法，包括研究CVSA下N2pc/ERP与SSVEP混合特征的基本诱发规律、基于ERP与SSVEP混合特征的超大指令高速BCI的范式设计、极微弱脑电特征的提取与识别算法、脑-机接口动态停止策略、系统集成与临床实验。本项目确定了CVSA下N2pc/ERP与SSVEP混合特征的基本诱发规律，建立了ERP的空、时、频、相多参数演进模型，筛选了可实现注意功能增强的SSVEP特征，构建了具有高信噪比ERP与SSVEP混合特征的诱发新方案，设计了高效的ERP与SSVEP混合编码策略，发展了基于空间滤波与模板匹配的极微弱脑电特征解码算法，建立了相应的BCI在线系统，实现信息传输率远超20 bits/min的传输速度，并已尝试将该系统用于基底动脉梗塞患者的日常交流。借助该BCI系统，患者能够表达出自己的意愿，如“想回家了”，“最想见的人是孙子”，“目前最不舒服的是腿部”等。在该项目的支持下，已发表或录用学术论文29篇，其中SCI期刊论文11篇，会议论文9篇，其中一篇被选为《IEEE transactions on Biomedical Engineering》封面论文，已申请专利4项。2018年，获首届Neuroscan Award奖项；2019年，获世界机器人大赛BCI脑控机器人比赛一等奖。