基于脑功能网络的抑郁脑电信号分类方法

Classification of Depression Electroencephalogram Signals Based on Brain Function Network

通过探究多种脑功能网络构建方法在抑郁脑电信号分类中的性能,提出了一种基于相位滞后指数、自适应阈值和卷积神经网络的高精度抑郁脑电信号分类方法。首先,利用不同连通性方法和二值化方法的组合构建了多种抑郁脑功能网络;然后,基于图论分析从脑功能网络中提取网络参数;最后,将脑网络参数输入多种分类器,实现抑郁脑电信号分类。此外,建立了一种基于可分性测度和量子粒子群优化的自适应阈值法和自适应密度法,以避免人工设置阈值和网络密度的主观性。实验结果表明:与其他常用方法相比,在alpha频带上,所提分类方法的性能最好,分类准确率和灵敏度分别为88.03%、89%。

By exploring the performance of brain function network construction methods,a high-precision classification method for depression electroencephalogram(EEG)was proposed.First,various depression brain function networks were constructed by using different connectivity and binarization methods.Subsequently,network parameters were extracted based on graph theory analysis.Finally,these parameters were input into various classifiers to achieve depression EEG classification.Additionally,an adaptive thresholding and density method based on divisibili-ty measure and quantum particle swarm optimization was established to avoid the subjectivity of manually setting thresholds and network density.Experimental results showed that in the alpha band,the classification method based on phase lag index,adaptive threshold,and convolutional neural network performs best with a classification accuracy of 88.03%and sensitivity of 89%.