基于多特征融合的睡眠分期

Sleep staging based on multi-feature fusion

目的为了有效实现睡眠自动分期,对睡眠障碍等相关疾病的诊断提供更多依据,本文提出了一种基于多特征融合的睡眠分期方法。方法数据来自ISRUC-Sleep数据库,首先对10名健康受试者和10名睡眠障碍患者的脑电(electroencephalogram,EEG)信号计算3种特征——样本熵、小波包能量和去趋势波动。然后采用支持向量机(support vector machine,SVM)构建睡眠分期模型,并验证该模型的准确性。此外,为了进行比较加入心电(electrocardiogram,ECG)和肌电(electromyogram,EMG)通道。结果健康受试者和睡眠障碍患者睡眠分期的准确率分别达到87.4%和86.3%。结论基于多特征融合的睡眠分期方法能够有效地提高睡眠分期的准确率。

Objective In order to realize the automatic sleep staging effectively and to provide more evidence for the diagnosis of sleep disorders and other related diseases,a sleep staging method based on multi-feature fusion is proposed in this paper.Methods Electroencephalogram(EEG)signals came from ISRUC-Sleep database,including 10 healthy subjects and 10 patients with sleep disorder problems.Three characteristics that include sample entropy,wavelet packet energy and detrended fluctuations were calculated.Then Support Vector Machine(SVM)was used to construct the sleep staging model and the accuracy of the model was verified.Additionally,electrocardiogram(ECG)and electromyogram(EMG)signals were also involved for comparison.Results The accuracy of sleep staging in two groups reached 87.4%and 86.3%respectively.Conclusions The sleep staging method based on multi-feature fusion can effectively improve the accuracy of sleep staging.