摘要
目前抑郁症的临床诊断多以医生经验和患者主观感受为主,主观性强、准确率低、耗时长。随着神经电生理学和计算机技术的发展,抑郁症的客观分类与识别成为可能。但是,已有的基于静息态脑电信号的抑郁症分类识别方法较为单一,脑电特征选取的精准性、综合性和有效性有待进一步探究。本文在设计包含两种模态实验范式的基础上,提出一种基于HFD和LZC特征联合的单通道静息态脑电抑郁症分类识别方法,以期用较少的特征获得较高的分类准确率。首先采集8名抑郁患者和8名健康对照的静息态脑电信号;然后提取其非线性动力学特征参数HFD和LZC;最后将特征数据输入到非线性支持向量机模型中进行分类识别。结果表明,联合特征得到的灵敏度、特异性和分类正确率最高分别为98.12%、96.67%和95.10%,较单独HFD/LZC特征平均分别提高了23.05%、17.02%和19.29%。同时,模型主体部分仅耗时约12 s。研究结果对临床实践中抑郁症的识别和辅助诊断具有重要意义。
At present, the clinical diagnosis of depression is mainly based on doctors′ experience and patients′ subjective feeling, which is highly subjective, low accuracy and time-consuming. With the development of neuron electrophysiology and computer technology, the objective classification and recognition of depression become possible. However, the existing research methods for the classification and identification of depression based on resting-state EEG signals are relatively simple, and it is necessary to further explore accurate, comprehensive and effective EEG features. In this article, a single-channel resting-state EEG depression classification and recognition method based on Higuchi′s Fractality Dimension(HFD) and Lempel-Ziv Complexity(LZC) is proposed based on the design of two experimental modes to obtain higher classification accuracy with fewer features. First, the resting-state EEG signals of 8 major depression disorders and 8 healthy control subjects are collected. Then, their nonlinear dynamic feature parameters HFD and LZC are extracted. Finally, the feature data are input into a nonlinear support vector machine model for classification recognition. Results show that the sensitivity, specificity and classification accuracy obtained by the combined feature are the highest at 98.12%, 96.67% and 95.10%, respectively, which are 23.05%, 17.02% and 19.29% higher than independent HFD/LZC. Meanwhile, the main part of the model only takes about 12 s. The findings have important implications for the identification and auxiliary diagnosis of depression in clinical practice.
作者
康显赟
刘爽
苏方玥
李洁
明东
Kang Xianyun;Liu Shuang;Su Fangyue;Li Jie;Ming Dong(Academy of Medical Engineering and Translational Medicine,Tianjin University,Tianjin 300072,China;Tianjin Mental Health Center,Tianjin 300222,China;College of Precision Instruments and Optoelectronics Engineering,Tianjin University,Tianjin 300072,China)
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2022年第7期181-190,共10页
Chinese Journal of Scientific Instrument
基金
国家杰出青年科学基金(81925020)
国家自然科学基金(81801786)
天津市自然科学基金(19JCYBJC29200)项目资助。
关键词
抑郁症
静息态脑电信号
脑电信号特征
分类识别模型
depression
resting-state EEG signals
features of EEG signal
classification and recognition model
