Abstract:

Based on the 64-channel EEG acquisition system of EGI Company, the EEG data of 16 adolescent patients with depression and 16 healthy patients with eyes closed for 4 minutes at resting-state data are collected. Secondly, Spectral Asymmetry Index（SASI） and Detrended Fluctuation Analysis（DFA） algorithm are applied to extract EEG time-frequency domain features. For the extracted features from the channel, on the hand, the features from the single optimal electrode Pz are selected as the channel of classification; on the other hand, the features from the multi-channels selected by Genetic Algorithm（GA） are used for classification. Finally, the Support Vector Machine（SVM） is applied to a single channel and multi-channels to classify depression patients and healthy people. The classification accuracies of the single channel are 45.5% and 51.5%, respectively, while the classification accuracies of the multi-channels selected by GA are 78.1% and 90.6%, respectively. Experimental results show that SASI algorithm is better than that of DFA real-time algorithm of computing, and the DFA algorithm accuracy is better than SASI algorithm. This study provides a theoretical basis for computer aided diagnosis of depression.

Key words: Electroencephalogram（EEG）, depression, Spectral Asymmetry Index（SASI）, Detrended Fluctuation Analysis（DFA）

摘要：

基于EGI公司64导脑电采集系统，采集了16位青少年抑郁症患者和16位正常人静息态下闭眼4分钟的脑电数据。运用频谱不对称分析法（Spectral Asymmetry Index，SASI）和去趋势波动分析（Detrended Fluctuation Analysis，DFA）算法提取脑电时域和频域特征。针对提取的特征的导联，一方面，选择最佳电极Pz作为分类的导联，另一方面，通过遗传算法对所有导联进行筛选，将筛选后的导联特征用于分类。使用支持向量机（Support Vector Machine，SVM）在单导联和多导联的情况下，对抑郁症患者和正常人进行分类，结果发现，单导联下，使用SVM分类器对抑郁组和对照组的SASI和DFA结果进行分类，分类精度分别为45.5%和51.5%，使用遗传算法的分类精度分别为78.1%和90.6%，SASI算法的计算实时性优于DFA算法，DFA算法的准确性优于SASI算法。该研究为抑郁症的计算机辅助诊断提供了理论依据。

关键词: 脑电信号（EEG）, 抑郁症, 频谱不对称分析（SASI）, 去趋势波动分析（DFA）