一种基于改进人工蜂群算法优化支持向量机的睡眠呼吸暂停检测方法

A sleep apnea detection method based on support vector machine optimized by improved artificial bee colony algorithm

目的睡眠呼吸暂停综合征(sleep apnea syndrome,SAS)是由于睡眠时上气道通气不畅或堵塞引起的呼吸暂停或低通气,严重影响人类健康和生活。目前的检测方法是多导睡眠仪,检测过程较为复杂,影响患者正常睡眠。为此本文提出了一种针对血氧饱和度信号的引入交叉变异的全局混沌人工蜂群(cross global chaos artificial bee colony,CGCABC)算法优化支持向量机(support vector machine,SVM)的SAS检测方法。方法从数据集ISRUC-SLEEP中提取25名SAS患者整晚8 h的脉搏血氧饱和度数据,经预处理后对每段数据计算5种非线性特征,包括近似熵、模糊熵、信息熵、排列熵和样本熵。比较发病片段信号特征和未发病片段信号特征之间的差异,使用CGCABC算法优化的SVM模型进行分类检测,并与人工蜂群(artificial bee colony,ABC)算法、粒子群(particle swarm optimization,PSO)算法、麻雀搜索(sparrow search,SS)算法优化SVM模型的检测结果进行对比。结果使用CGCABC算法优化的SVM模型在准确率、特异度、敏感度以及收敛时间上均有较好的效果,优于ABC算法、PSO算法和SS算法优化SVM模型的检测。结论本文提出的方法对SAS这一疾病的识别和检测具有重要价值,在医疗领域上具有广泛的应用前景。

Objective Sleep apnea syndrome(SAS)is apnea or hypopnea caused by poor or blocked upper airway ventilation during sleep,which seriously affects human health and life.At present,the detection method is polysomnography,whose process is so complicated that affects the normal sleep of patients.Therefore,this paper proposes a SAS detection method based on the support vector machine(SVM)optimized by the cross global chaos artificial bee colony(CGCABC)algorithm for blood oxygen saturation signal.Methods The pulse oximetry data of 25 SAS patients were extracted from ISRUC-SLEEP dataset for 8 hours overnight.After pre-processing,five nonlinear features were calculated for each segment of data,including approximate entropy,fuzzy entropy,information entropy,permutation entropy and sample entropy.Results Comparing the detection results of SVM model optimized by the algorithms of artificial bee colony(ABC),particle swarm optimization(PSO)and sparrow search(SS),CGCABC-SVM model had good results in accuracy,specificity,sensitivity and running time.Conclusions The method proposed in this paper has important value on SAS identification and detection,which has a wide application prospect in the medical field.