東華大學圖書館 |

Statistical Methods for Wearable Device Data: Applications in Clinical Studies = = 可穿戴设备数据的统计分析方法及其在临床研究中的应用.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Statistical Methods for Wearable Device Data: Applications in Clinical Studies =/
其他題名:	可穿戴设备数据的统计分析方法及其在临床研究中的应用.
作者:	Li, Xinyue.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2019,
面頁冊數:	99 p.
附註:	Source: Dissertations Abstracts International, Volume: 81-03, Section: B.
Contained By:	Dissertations Abstracts International81-03B.
標題:	Biostatistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13807152
ISBN:	9781085777438

Statistical Methods for Wearable Device Data: Applications in Clinical Studies = = 可穿戴设备数据的统计分析方法及其在临床研究中的应用.
Li, Xinyue.

Statistical Methods for Wearable Device Data: Applications in Clinical Studies =可穿戴设备数据的统计分析方法及其在临床研究中的应用. - Ann Arbor : ProQuest Dissertations & Theses, 2019 - 99 p.

Source: Dissertations Abstracts International, Volume: 81-03, Section: B.

Thesis (Ph.D.)--Yale University, 2019.

This item must not be sold to any third party vendors.

Wearable devices have been increasingly used in clinical research to provide continuous and objective measures of physical activity and further study sleep, activity, and circadian rhythms. Wearable devices avoid the issues with traditional sleep and activity logs that are limited by subjectivity, bias and extra manual work for caregivers, and it is also low-cost and easy to wear. However, the analysis of time series data from actigraphy remains the major obstacle for researchers and there are few statistical methods tailored to analyze and interpret the time-series accelerometer data. Therefore, it is imperative to develop statistical methodologies and computationally efficient tools to analyze the data and translate the rich information into biological meaningful insights. In this thesis, we have developed novel statistical methods for sleep/wake identification and circadian rhythm analysis as well as efficient computational tools for method implementation. Applications to early childhood clinical studies and large-scale population studies yield novel insights into early childhood physical development and the underlying genetic architectures of sleep and activity with physical and mental health traits.Sleep/wake identification is crucial for sleep research, as accurately defined sleep start, sleep end and sleep duration are critical for downstream analysis. Current sleep/wake identification algorithms often require the collection of polysomnography, the "gold-standard" in sleep research, to train the model, but polysomnography has the disadvantages of high costs, in-lab setting, intrusive measures, and difficulty in long-time monitoring. Current methods are also labor-intensive in model training steps, arbitrary in variable selection or threshold setting, and ad-hoc in the limited use of each trained algorithm in one dataset. We propose a Hidden Markov Model based algorithm that automatically categorizes epochs into sleep/wake states without human supervision. The proposed HMM-based algorithm is unsupervised that saves manual work and is also data-driven and directly applicable to data from different sources. The unsupervised algorithm can expand the application of actigraphy in large epidemiologic studies as well as in cases where intrusive polysomnography is hard to use, such as in pediatric populations. As an added benefit, the estimated HMM parameters can capture individual variabilities in sleep and activity patterns and one can use the information for further analysis.For circadian rhythm analysis, current parametric methods often assume a sinusoidal shape of activity curves that are often violated. Current nonparametric methods extracting simple metrics such as mean activity levels during five most active hours and ten least active hours of the day cannot make full use of the rich information contained in data. None of current statistical methods effectively analyze the periodic information in sleep-wake circadian rhythms, even though circadian rhythms are characterized by periodicities. We propose a Penalized Multi-band Learning approach that can analyze actigraphy to select dominant periodicities sequentially and further use periodic information to characterize the sleep-wake circadian rhythm of the study population. Application of our method to an early childhood dataset provides new insights into early childhood development by finding the association between circadian rhythm formation and motor development, which was not identified in previous studies.Wearable devices make it possible to achieve long-term activity monitoring of a large population, but the analysis of large and complex accelerometer data remains the major challenge. The situation will be even complicated when combining wearable device data with complex genetic data to conduct genetic research. We have developed efficient computational tools to implement the proposed statistical methods described above and applied them to the UK Biobank study, which contains accelerometer data from ~100,000 participants and genotyping data from ~500,000 participants. Sleep, activity, and circadian features were extracted from accelerometer data and 18 significantly associated loci were identified, 14 of which are novel. Our study demonstrates the utility of wearable devices in studying sleep and circadian rhythms in genetic studies and large-scale population studies and provides novel insights into the shared genetic architectures of sleep and activity with physical, mental and neurological traits. Overall, the proposed statistical methods can provide efficient tools to decipher accelerometer data and help better understand the association of sleep, activity, circadian rhythms with disease etiology.

ISBN: 9781085777438Subjects--Topical Terms:

1002712
Biostatistics.
Subjects--Index Terms:

Accelerometer

Statistical Methods for Wearable Device Data: Applications in Clinical Studies = = 可穿戴设备数据的统计分析方法及其在临床研究中的应用.
LDR:05974nmm a2200373 4500 001 2272387
005 20201105110043.5
008 220629s2019 ||||||||||||||||| ||eng d
020 $a 9781085777438
035 $a (MiAaPQ)AAI13807152
035 $a AAI13807152
040 $a MiAaPQ $c MiAaPQ
100 1 $a Li, Xinyue. $3 3549822
245 1 0 $a Statistical Methods for Wearable Device Data: Applications in Clinical Studies = $b 可穿戴设备数据的统计分析方法及其在临床研究中的应用.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2019
300 $a 99 p.
500 $a Source: Dissertations Abstracts International, Volume: 81-03, Section: B.
500 $a Advisor: Zhao, Hongyu.
502 $a Thesis (Ph.D.)--Yale University, 2019.
506 $a This item must not be sold to any third party vendors.
520 $a Wearable devices have been increasingly used in clinical research to provide continuous and objective measures of physical activity and further study sleep, activity, and circadian rhythms. Wearable devices avoid the issues with traditional sleep and activity logs that are limited by subjectivity, bias and extra manual work for caregivers, and it is also low-cost and easy to wear. However, the analysis of time series data from actigraphy remains the major obstacle for researchers and there are few statistical methods tailored to analyze and interpret the time-series accelerometer data. Therefore, it is imperative to develop statistical methodologies and computationally efficient tools to analyze the data and translate the rich information into biological meaningful insights. In this thesis, we have developed novel statistical methods for sleep/wake identification and circadian rhythm analysis as well as efficient computational tools for method implementation. Applications to early childhood clinical studies and large-scale population studies yield novel insights into early childhood physical development and the underlying genetic architectures of sleep and activity with physical and mental health traits.Sleep/wake identification is crucial for sleep research, as accurately defined sleep start, sleep end and sleep duration are critical for downstream analysis. Current sleep/wake identification algorithms often require the collection of polysomnography, the "gold-standard" in sleep research, to train the model, but polysomnography has the disadvantages of high costs, in-lab setting, intrusive measures, and difficulty in long-time monitoring. Current methods are also labor-intensive in model training steps, arbitrary in variable selection or threshold setting, and ad-hoc in the limited use of each trained algorithm in one dataset. We propose a Hidden Markov Model based algorithm that automatically categorizes epochs into sleep/wake states without human supervision. The proposed HMM-based algorithm is unsupervised that saves manual work and is also data-driven and directly applicable to data from different sources. The unsupervised algorithm can expand the application of actigraphy in large epidemiologic studies as well as in cases where intrusive polysomnography is hard to use, such as in pediatric populations. As an added benefit, the estimated HMM parameters can capture individual variabilities in sleep and activity patterns and one can use the information for further analysis.For circadian rhythm analysis, current parametric methods often assume a sinusoidal shape of activity curves that are often violated. Current nonparametric methods extracting simple metrics such as mean activity levels during five most active hours and ten least active hours of the day cannot make full use of the rich information contained in data. None of current statistical methods effectively analyze the periodic information in sleep-wake circadian rhythms, even though circadian rhythms are characterized by periodicities. We propose a Penalized Multi-band Learning approach that can analyze actigraphy to select dominant periodicities sequentially and further use periodic information to characterize the sleep-wake circadian rhythm of the study population. Application of our method to an early childhood dataset provides new insights into early childhood development by finding the association between circadian rhythm formation and motor development, which was not identified in previous studies.Wearable devices make it possible to achieve long-term activity monitoring of a large population, but the analysis of large and complex accelerometer data remains the major challenge. The situation will be even complicated when combining wearable device data with complex genetic data to conduct genetic research. We have developed efficient computational tools to implement the proposed statistical methods described above and applied them to the UK Biobank study, which contains accelerometer data from ~100,000 participants and genotyping data from ~500,000 participants. Sleep, activity, and circadian features were extracted from accelerometer data and 18 significantly associated loci were identified, 14 of which are novel. Our study demonstrates the utility of wearable devices in studying sleep and circadian rhythms in genetic studies and large-scale population studies and provides novel insights into the shared genetic architectures of sleep and activity with physical, mental and neurological traits. Overall, the proposed statistical methods can provide efficient tools to decipher accelerometer data and help better understand the association of sleep, activity, circadian rhythms with disease etiology.
590 $a School code: 0265.
650 4 $a Biostatistics. $3 1002712
650 4 $a Statistics. $3 517247
653 $a Accelerometer
653 $a Actigraphy
653 $a Methodology
653 $a Physical activity
653 $a Statistics
653 $a Wearable device
690 $a 0308
690 $a 0463
710 2 $a Yale University. $b Statistics and Data Science. $3 3549823
773 0 $t Dissertations Abstracts International $g 81-03B.
790 $a 0265
791 $a Ph.D.
792 $a 2019
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=13807152