航天医学与医学工程

目的 探索腕带获取的血容量脉搏及体动三轴加速度在睡眠分期中的作用。方法 使用DREAMT公开数据库全部100例睡眠障碍者的Empatica E4腕带血容量脉搏(BVP)及体动三轴加速度(ACC)，使用BVP基线的两个频域特征[频域能量峰占比(e_s)和低频能量/高频能量(LF/HF)]、一个时域特征[幅度变异性(v_A)]及ACC的活动计数(C_s)，基于随机森林进行睡眠、觉醒两分类。结果 通过“留一法”交叉验证获得的全部100例睡眠障碍者的睡眠、觉醒两分类结果：基于BVP及ACC共4个特征的两分类精度为79.8%,Kappa系数为0.56；基于BVP的3个特征的两分类精度为70.4%,Kappa系数为0.36；基于ACC的两分类精度为75.1%,Kappa系数为0.47。结论 腕带采集的血容量脉搏及体动三轴加速度可用于对睡眠障碍者睡眠、觉醒的粗略估计，其中体动三轴加速度的重要性高于血容量脉搏。

Objective To explore the role in sleep staging from blood volume pulse(BVP) and triaxial acceleration(ACC) of body movement obtained by wristband.Methods The BVP and ACC obtained by Empatica E4 wristband were used from all 100 cases of sleep disorder subjects in the DREAMT public database.Two frequency domain characteristics(e_S,LF/HF) and one time domain characteristic(v_A) of the BVP baseline and the activity counts(C_S) of the ACC were used for sleep-awakening classification based on random forest.Results The results of sleep-awakening classification of all 100 cases of sleep disorder subjects were obtained by leaving-one-out strategy.The accuracy is 79.8% and the Kappa coefficient is 0.56 by 4 features from BVP and ACC;the accuracy is 70.4% and the Kappa coefficient is 0.36 by 3 features of BVP;the accuracy is 75.1% and the Kappa coefficient is 0.47 based on activity counts.Conclusion The BVP and ACC obtained by the wristband can be used for the rough estimation of sleep and awakening for sleep disorder subjects,among which the importance of ACC is higher than that of BVP.