基于心率变异性与机器学习的睡眠呼吸事件分类  

Classification of sleep respiratory events based on heart rate variability and machine learning

作　　者：梁九兴[1] 张湘民[2] 黄少雄[1] 曾令紫 罗语溪[1] LIANG Jiuxing;ZHANG Xiangmin;HUANG Shaoxiong;ZENG Lingzi;LUO Yuxi(School of Engineerng,Sun Yat-sen University,Guangzhou 510006,China;The Sixth Affiliated Hospital of Sun Yat-sen U n iversity,Guangzhou 510655,（china)

机构地区：[1]中山大学工学院,广东广州510006 [2]中山大学附属第六医院,广东广州510655

出　　处：《中山大学学报（自然科学版）》2018年第3期128-134,共7页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：国家自然科学基金(81570904);广东省自然科学基金(2014A030313215);广东省科技计划项目(2017B020210007)

摘　　要：为了提供一种针对睡眠呼吸暂停低通气综合征(sleep apnea hypopnea syndrome,SAHS)患者的筛查方法,本研究把心率变异性(heart rate variability,HRV)应用于睡眠呼吸模态的分类问题。通过构建和训练概率神经网络(probabilistic neural network,PNN)对HRV各特征值进行有无异常睡眠呼吸事件的判别,以期实现对该病征进行初步筛查的目的。首先,对标注的有无呼吸事件的多导睡眠监测数据提取其心电的HRV特征值,再经过归一化后作为特征向量;其次采用PNN分类算法对特征向量进行训练及分类输出;最后,对模型的预测分类性能进行评价。对于准确率、灵敏度、特异性、受试者工作特性曲线下面积(area under the receiver operating characteristic curve,AUC)及分类耗时等评价指标PNN分类器的结果分别为:75.97%,82.51%,76.22%,0.7936,0.63 s。与广义回归神经网络(generalized regression neural network,GRNN)及极限学习机(extreme learning machine,ELM)分类算法相比,PNN分类算法在灵敏度、特异性、AUC及分类耗时评价维度上均取得最优。本研究基于HRV及PNN分类算法实现了对有无异常睡眠呼吸事件的判别,提供了一种低生理负荷SAHS筛查的途径。To provide a method for screening patients with sleep apnea hypopnea syndrome （ SAHS ）,the heart rate variability （ HRV） was applied to the classification of sleep respirator neural net＂work （ PNN） was proposed to classify the normal and abnormal sleep respiratory eventsaccording to the HRV features to achieve the purpose of preliminary screening of the disease. In this classification process,the HRV features of ECG were firstly extracted from the polysomnographic monitorngdata related to the normal and abnormal sleep respiratory events,and then normalized Then,PNN classification algorithm was used to train and classify the features. The prediction and classification performance of the model was finally evaluated. The results of the PNN classifier cy,sensitivity,speci f ici ty,area under the receiver operating characteristic curve （ AUC ） o f the subjects and time consumption for classif icat ion were re spective ly： 75. 9 7 % , 82. 5 1 % , 76. 22 % , 0. 7936 and0. 63 s. Compared wi th generalized regression neural network （ GRN N ） and extreme learning machine （ELM ） classification algorithms , PNN classification algorithm is opt imal in sensit ivity , spe c if ic ity , AUC and time consumptions. In this study , HRV and PNN classif icat ion algorithm were used to classify the presence or absence of abnormal sleep respiratory events , thus provid in g a low physiological load SAHS screening method. The study has a certain pra c tical significance for the initial screening of the disease.

关 键 词：心率变异性 睡眠呼吸事件 机器学习 分类算法 

分 类 号：R318[医药卫生—生物医学工程]