穿戴式裝置睡眠與活動參數之探索性研究：以UKB資料庫為例

An Exploratory Analysis of Sleep and Activity Variables from Wearable Device in UK Biobank Study

蘇千蘅(Chian-Heng Su)

指導教授 ： 蕭朱杏

共同指導教授 ： 杜裕康(Yu-Kang Tu)

國立臺灣大學/公共衛生學院/流行病學與預防醫學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202200492

摘要

腕部穿戴式裝置是近年常被用來記錄身體活動的媒介，它能記錄在X、Y、Z三軸加速度的時序資料以偵測手部活動。然而，要分析這類型資料，有一些問題需要考量，例如這個資料的可信度、及如何把這個資料與其他型態的資料結合；另外，透過穿戴式裝置資料在不同表型或疾病的相關性分析，能讓這些資料有機會應用於實際生活中。本研究作為前驅研究，希望能了解穿戴式裝置資料的特質、與其他資料變數的關係，以及作為後續分析變數的可行性。研究以英國生物資料庫的23,920人為樣本，聚焦於睡眠區間內的穿戴式裝置資料，經由R軟體GGIR套件和HDCZA演算法完成主要的資料前處理以及睡眠區間與參數的估計，比較問卷與穿戴式裝置兩種不同測量方式所收集到的睡眠參數之間的共通性；並以資料庫中提供之精神狀態為例，比較重度憂鬱、躁鬱和健康對照組之間Angle Z手部活動函數的不同。研究結果發現，在英國生物資料庫中，受試者的睡眠時長和晝夜節律不因收案方式和收案期間的不同而出現太大的改變；基於問卷與穿戴式裝置資料之間所表現的共通性，在沒有睡眠標準測量PSG紀錄的情況下，增加了這些睡眠測量結果的可信度。基於這個發現，合併兩種資料後分析顯示，若針對入睡後前三小時的手部活動角度(Angle Z)而言，平均Angle Z的差異主要來自性別和肥胖程度，而個人Angle Z的變異則主要和性別、年齡、肥胖程度、打呼和睡眠時長固定與否有關。穿戴式裝置資料量雖然龐大且須經過複雜的前處理，然而在需要連續持續記錄的參數、尤其是客觀參數時，仍有其優勢。不過，因為穿戴式裝置難以捕捉主觀或情感面之特徵參數，建議未來研究者使用這樣的樣本資料尋找疾病相關性時，須注意其限制。

關鍵字

穿戴式裝置 ； 腕部穿戴式裝置 ； 睡眠特徵 ； 憂鬱或躁鬱狀態 ； 英國生物資料庫 ； R軟體套件GGIR

並列摘要

Wrist-worn accelerometer takes records of acceleration in three axes to denote arm movement. Issues such as the validity and comparability with other data arise however when this type of records is used for analysis. With these wearable device data in analysis, scientists may be able to solve real life problems. This research is a pilot study aiming to understand its properties, relationship to other variables, and the potential to be combined with other data in future studies, based on the data retrieved from 23,920 subjects wearing accelerometers in UK Biobank project. The R package GGIR and the Heuristic algorithm looking at Distribution of Change in Z-Angle (HDCZA) algorithm were used for pre-processing the acceleration records to obtain values of Angle Z and sleep parameters including sleep period time (SPT) window. The results were then compared with responses collected in the UKB sleep questionnaire to evaluate the compatibility between these two types of data. In addition, this research compares the difference in Angle Z between subjects of different mental status. Our first finding is that the sleep duration and chronotype estimates from questionnaire and accelerometer were similar, even though the data collection duration of questionnaire was 3 to 6.5 years prior to the data collection with accelerometer. It implies that the sleep measurements from the wearable device were comparable to that collected by questionnaire. These two sets of parameters were then combined to examine if the mean Angle Z in the first three hours of sleep varies between different phenotypic groups. The results indicate that the mean Angle Z is associated with gender and obesity, and the subject-specific variation in Angle Z was found associated with gender, age, obesity, snoring and the consistency of sleep duration. Although the wearable device can provide large amount of data and the pre-processing of such data has been standardized, it has the limitation in collecting subjective features. Research studies involving such variables and accelerometers should be dealt with care.

並列關鍵字

wearable device ； wrist-worn accelerometer ； sleep parameters ； bipolar and major depression status ； UK Biobank ； GGIR

參考文獻

1. Hale, L., W. Troxel, and D.J. Buysse, Sleep Health: An Opportunity for Public Health to Address Health Equity. Annual Review of Public Health, 2020. 41(1).

2. Besedovsky, L., T. Lange, and J. Born, Sleep and immune function. Pflügers Archiv - European Journal of Physiology, 2012. 463(1).

3. Rasch, B. and J. Born, About sleep's role in memory. Physiol Rev, 2013. 93(2).

4. Buysse, D.J., Sleep health: can we define it? Does it matter? Sleep, 2014. 37(1).

5. Kyle, S.D., et al., Sleep and cognitive performance: cross-sectional associations in the UK Biobank. Sleep Med, 2017. 38.