基于脑电功能连接拓扑表征的心算任务分类

使用脑网络图的方法分析脑电功能连接存在阈值选择、忽略了脑电图动力学特性的问题。针对这一问题，提出了一种使用拓扑动态建模的方法来分析脑电功能连接矩阵，以提高心算任务分类识别正确率。该方法首先将功能连接矩阵转换为无向加权图，然后使用持续同调工具来构建不同的复形，记录拓扑动态过程中形成的不同阶的同调特征，形成持续图，最后使用持续景观图特征作为分类特征，输入到随机森林分类器进行心算状态识别。在心算状态识别和心算质量分类两个任务中分别获得了最高99.26%、99.20%的识别准确率，97.87%、99.80%的敏感性，以及99.78%、97.64%的特异性，并且在跨个体验证实验中分别获得了66.81%、66.85%的准确率。实验结果表明，该方法能充分考虑所有可能的阈值，有效提取脑电功能连接的分类信息，实现脑电心算状态自动识别。

Mental arithmetic task classification based on topological representation of EEG-based functional connectivity

Using the method of brain network graph to analyze EEG functional connectivity has the problems of threshold selection and ignoring brain dynamic. To solve this problem, This paper proposed a method of using topological dynamic mode-ling to analyze the EEG functional connectivity matrix, which improved the accuracy of classification and recognition of mental arithmetic tasks. Firstly, it mapped the functional connectivity matrix to an undirected weighted graph. Then used the persistent Homology Toolbox to construct different complexes and record the different levels of homology features formed in this topological dynamic process to form the persistence diagrams. Finally, it calculated the persistence landscape features as the input feature of the random forest classifier for mental state recognition. In the two task of mental arithmetic state recognition and mental arithmetic quality classification, the proposed algorithm obtained the highest recognition accuracy at 99.26% and 99.20%, sensitivity at 97.87% and 99.80% and specificity at 99.78% and 97.64%, respectively, and accuracy at 66.81% and 66.85% in the cross-individual verification experiment. Experimental results showed that the proposed algorithm was fully considered all possible thresholds and effectively extracted the classification information of EEG functional connectivity to implement the automatic recognition of EEG mental arithmetic state.