北京生物医学工程

基于脑电信号及共空间模式的抑郁症多分类算法研究

Research on multiclass classification algorithm ofdepressionbased on EEG signal and common spatial pattern

作者：杨子贤罗涛李剑峰范艺晶

单位：北京邮电大学（北京 100876） 通信作者：罗涛。E-mail：tluo@bupt.edu.cn

分类号：R318.04

出版年·卷·期（页码）：2022·41·2（167-173）

摘要：

目的为了实现抑郁症患者的早发现、早治疗，提高抑郁症识别准确率及轻度患者召回率，对抑郁症的人工智能辅助诊断方法进行研究，提出了一种基于共空间模式的脑电信号二分类器投票算法。方法首先，提出一种基于方差分析确定最佳数据分段时长的方法进行数据预处理，并引入共空间模式算法进行特征提取。然后，结合当前二分类学习器解决多分类问题的扩展框架，设计了一种二分类器投票算法，解决了共空间模式算法在多分类问题上的局限性，实现健康对照、轻度患者、重度患者的脑电信号三分类。最后，利用抑郁症研究公开数据集MODMA，以准确率及轻度患者召回率为评价指标，通过对比实验验证算法在性能上的优势。结果在公开数据集上，基于共空间模式的投票算法准确率高达97.55%，轻度患者召回率为91%，与两种传统的多分类共空间模式扩展策略性能相比，准确率分别提升1.32%和5.10%，轻度患者召回率分别提升9%和18%。结论基于共空间模式的投票算法能有效提高抑郁症识别准确率及轻度患者召回率，可为促进抑郁症的早发现早治疗提供算法支持。

Objective In order to realize the early detection and treatment of patients with depression, improve the recognition accuracy of depression and the recall rate of mild patients, the artificial intelligence aided diagnosis method of depression is studied, and a voting algorithm based on two-classifiers of EEG signal is proposed. Methods Firstly, a method based on variance analysis to determine the optimal data segmentation time is proposed for data preprocessing, and the common space pattern algorithm is introduced for feature extraction. Then, combined with the extended framework of the current two-classifiers to solve the multi classification problem, a two-classifiers voting algorithm is designed to solve the limitation of the common spatial pattern algorithm in the multi classification problem, and realize the three classification of EEG signals of healthy controls, mild patients and severe patients. Finally, using the open data set MODMA of depression research, taking the accuracy rate and mild patient recall rate as evaluation indexes, the performance advantages of the algorithm are verified through comparative experiments. Results On the public dataset, the accuracy rate of the voting algorithm based on the common spatial pattern is as high as 97.55%, and the mild patient recall rate is 91%. Compared with the performance of the two traditional multi classification common spatial pattern extension strategies, the accuracy rate is improved by 1.32% and 5.10%, and the mild patient recall rate is improved by 9% and 18%. Conclusions The voting algorithm based on the common space pattern can effectively improve the recognition accuracy of depression and the recall rate of mild patients, and can provide algorithm support for promoting the early detection and treatment of depression.

参考文献：

[1]Vakalopoulos?C. The EEG as an index of neuromodulator balance in memory and mental illness[J]. Frontiers in Neuroscience, 2014, 8:?63.

[2]Carney PW, Masterton R,?Flanagan?D, et al. The frontal lobe in absence epilepsy: EEG-fMRI findings[J]. Neurology, 2012, 78(15): 1157-1165.

[3]Morabito?FC, Labate?D, Foresta FL, et al. Multivariate multi-scale permutation entropy for complexity analysis of Alzheimer's disease EEG[J]. Entropy, 2012, 14: 1186-1202.

[4]Schmidt MT, Kanda P, Basile L, et al. Index of Alpha/theta ratio of the electroencephalogram: a new marker for Alzheimer's disease[J]. Frontiers in Aging Neuroscience, 2013, 5: 60.

[5]Lebowitz M. Integrated learning: controlling explanation[J]. Cognitive Science, 1986, 10(2): 219-240.

[6]Yu C, Li J,?Zhao?F, et al. Comparative study on face recognition based on SVM of one-against-one and one-against-rest methods[C]//?2014 8th?International Conference on Future Generation Communication & Networking.?Hainan Island,China:?IEEE?Press, 2014: 104-107.

[7]Mohammadi Y, Hajian M, Moradi MH. Discrimination of depression levels using machine learning methods on EEG signals[C]//?2019 27th Iranian Conference on Electrical Engineering (ICEE). Yazd,Iran: ICEE,?2019: 1765-1769.

[8]Cai H, Han J, Chen Y, et al. A pervasive approach to EEG-based depression detection[J]. Complexity, 2018, 2018: 5238028.

[9]Bachmann?M, P?eske?L, Kalev K, et al. Methods for classifying depression in single channel EEG using linear and nonlinear signal analysis[J]. Computer Methods and?Programs in Biomedicine, 2018, 155: 11-17.

[10]喇蓉. 基于脑电信号和深度神经网络的轻度抑郁识别研究[D].?兰州: 兰州大学,?2020.La R. Mild depression recognition based on EEG and deep neural network[D]. ?Lanzhou: Lanzhou University, 2020.

[11]Henriques JB, Davidson?RJ. Regional brain electrical asymmetries discriminate between previously depressed and healthy control subjects[J]. Journal of Abnormal Psychology, 1990, 99(1):?22-31.

[12]Grin-Yatsenko VA, Baas I, Ponomarev VA, et al. EEG power spectra at early stages of depressive disorders[J]. Journal of Clinical Neurophysiology:?Official Publication of the American Electroencephalographic Society, 2009, 26(6):?401-406.

[13]Koles ZJ, Lazar MS, Zhou SZ. Spatial patterns underlying population differences in the background EEG[J]. Brain Topography, 1990, 2(4):?275-284.

[14]王建军, 李乐加. 截取不同长度数据段对脑电地形图结果的影响[J]. 临床脑电学杂志, 1995,?4(27):?241-242.??????

[15]李立亭. 基于共空间模式的四类运动想象脑电信号分类方法[J]. 仪器仪表用户, 2016,?23(5): 12-14.Li LT. Classification method of four-class motor imagery EEG data based on common spatial pattern[J]. Instrumentation Customer, 2016,?23(5): 12-14.

[16]Cai HS, Gao YW, Sun ST,?et al.?MODMA dataset: a multi-modal open dataset for mental-disorder analysis[EB/OL].??(2020-02-20)[2022-01-17]. https://dblp.org/rec/journals/corr/abs-2002-09283.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录