Detecting Sleep and Nonwear in 24-h Wrist Accelerometer Data from the National Health and Nutrition Examination Survey.
Detecting Sleep and Nonwear in 24-h Wrist Accelerometer Data from the National Health and Nutrition Examination Survey.
复制标题
DOI：
10.1249/mss.0000000000002973
复制
发表时间：
2022-11-01
影响因子：
4.1
通讯作者：
Intille S
中科院分区：
文献类型：
--
作者：
Thapa-Chhetry B;Arguello DJ;John D;Intille S
从腕戴式加速度计数据估计身体活动、久坐行为和睡眠需要在睡眠和清醒期间可靠地检测传感器非磨损和传感器磨损。开发一种算法，同时识别传感器唤醒磨损，睡眠磨损，和非磨损在24小时手腕加速度计数据收集与过滤或不过滤。使用标记有多导睡眠图的传感器数据（N=21）和直接观察到的健康成年人的醒着数据（N=31），以及来自留在家中各个位置的传感器的无磨损数据（N=20），我们开发了一种算法来检测无磨损、睡衣、以及2011-2014年国家健康和营养调查中收集的“空闲睡眠模式”（ISM）过滤数据的唤醒磨损。然后，该算法被扩展到处理从没有ISM过滤的设备收集的原始数据。使用基于多导睡眠图的睡眠和清醒磨损数据集（N=22）和来自健康成人的基于日记的清醒磨损和非磨损标签（N=23）进一步验证了这两种算法。将分类性能（F1分数）与四种替代方法进行了比较。基于ISM的算法在训练数据集上使用留一受试者交叉验证的F1得分为0.95±0.13。对两个独立数据集的验证得出的F1评分为0.84±0.60（睡眠时穿戴和清醒时穿戴的数据集）和0.94±0.04（清醒时穿戴和不穿戴的数据集）。使用原始原始数据时，训练数据集的F1评分为0.96±0.08，两个独立验证数据集的F1评分为0.86±0.18和0.97±0.04。该算法在数据集上的性能优于或优于其他方法。设计了一种新的机器学习算法，用于识别24小时腕戴式加速度计数据中的唤醒磨损、睡眠磨损和非磨损，该算法适用于ISM过滤数据或原始原始数据。
Estimating physical activity, sedentary behavior, and sleep from wrist-worn accelerometer data requires reliable detection of sensor non-wear and sensor wear during both sleep and wake. To develop an algorithm that simultaneously identifies sensor wake-wear, sleep-wear, and non-wear in 24-hour wrist-accelerometer data collected with or without filtering. Using sensor data labeled with polysomnography (N=21) and directly observed wake-wear data (N=31) from healthy adults, and non-wear data from sensors left at various locations in a home (N=20), we developed an algorithm to detect non-wear, sleep-wear, and wake-wear for ‘idle sleep mode’ (ISM) filtered data collected in the 2011–2014 National Health and Nutrition Examination Survey. The algorithm was then extended to process original raw data collected from devices without ISM filtering. Both algorithms were further validated using a polysomnography-based sleep and wake-wear dataset (N=22) and diary-based wake-wear and non-wear labels from healthy adults (N=23). Classification performance (F1-scores) was compared to four alternative approaches. F1-score of the ISM-based algorithm on the training dataset using leave-one-subject-out cross-validation was 0.95±0.13. Validation on the two independent datasets yielded F1-scores of 0.84±0.60 for the dataset with sleep-wear and wake-wear and 0.94±0.04 for the dataset with wake-wear and non-wear. F1 score when using original, raw data was 0.96±0.08 for the training datasets and 0.86±0.18 and 0.97±0.04 for the two independent validation datasets. The algorithm performed comparably or better than the alternative approaches on the datasets. A novel machine-learning algorithm was designed to recognize wake-wear, sleep-wear, and non-wear in 24-hour wrist-worn accelerometer data that is applicable for ISM-filtered data or original raw data.