Roach Gregory D, Miller Dean J, Shell Stephanie J, Miles Kathleen H, Sargent Charli
Appleton Institute for Behavioural Science, Central Queensland University, Adelaide 4701, Australia.
Australian Institute of Sport, Australian Sports Commission, Canberra 2617, Australia.
Sensors (Basel). 2025 Mar 27;25(7):2123. doi: 10.3390/s25072123.
The aim of the study was to examine the validity of a neurophysiological-based wearable device, i.e., Somfit (Compumedics Ltd.), for the assessment of sleep in athletes. Twenty-seven athletes (14 F, 13 M, aged 22.3 ± 5.1 years) spent a single night in a sleep laboratory. The participants had 9 h in bed (23:00-08:00) while fitted simultaneously with Somfit and polysomnography (PSG), i.e., the gold standard for the assessment of sleep. Somfit and PSG were used to independently categorise each 30-s epoch of time in bed into one of five states, i.e., wake, stage 1 non-REM sleep (N1), stage 2 non-REM sleep (N2), stage 3 non-REM sleep (N3), or REM sleep. There were large differences between participants in terms of the amount of Somfit data that were successfully captured/scored, so three subsets were considered in the subsequent analyses: unfiltered subset (n = 26)-all participants, except one for whom no Somfit data were captured/scored; good-capture subset (n = 15)-participants for whom > 80% of Somfit data were captured/scored; excellent-capture subset (n = 7)-participants for whom > 99.9% of Somfit data were captured/scored. Agreement for the five-state categorisation of time in bed was calculated as the percentage of PSG epochs correctly scored by Somfit as N1, N2, N3, REM, or wake. Agreement (and Cohen's kappa) was 63% (0.47) for the unfiltered subset, 66% (0.52) for the good-capture subset, and 79% (0.70) for the excellent-capture subset. These data indicate a moderate-substantial level of agreement between Somfit and PSG for the assessment of sleep in athletes. Wearable devices that can capture valid sleep data may also be used to derive important measures related to the circadian system, such as sleep consistency and social jet lag.
本研究的目的是检验一种基于神经生理学的可穿戴设备，即Somfit（Compumedics有限公司）在评估运动员睡眠方面的有效性。27名运动员（14名女性，13名男性，年龄22.3±5.1岁）在睡眠实验室度过一晚。参与者在床上躺9小时（23:00至08:00），同时佩戴Somfit和多导睡眠图（PSG），即评估睡眠的金标准。Somfit和PSG被用于独立地将在床上的每30秒时间段分类为五种状态之一，即清醒、非快速眼动睡眠1期（N1）、非快速眼动睡眠2期（N2）、非快速眼动睡眠3期（N3）或快速眼动睡眠。在成功捕获/评分的Somfit数据量方面，参与者之间存在很大差异，因此在后续分析中考虑了三个子集：未过滤子集（n=26）——所有参与者，除了一名未捕获/评分到Somfit数据的参与者；良好捕获子集（n=15）——捕获/评分到超过80%的Somfit数据的参与者；优秀捕获子集（n=7）——捕获/评分到超过99.9%的Somfit数据的参与者。床上时间五状态分类的一致性通过Somfit正确分类为N1、N2、N3、快速眼动或清醒的PSG时间段的百分比来计算。未过滤子集的一致性（和科恩kappa系数）为63%（0.47），良好捕获子集为66%（0.52），优秀捕获子集为79%（0.70）。这些数据表明，在评估运动员睡眠方面，Somfit和PSG之间存在中等至高度的一致性。能够捕获有效睡眠数据的可穿戴设备也可用于得出与昼夜节律系统相关的重要指标，如睡眠一致性和社会时差。
