Validation of the Munich Actimetry Sleep Detection Algorithm for estimating sleep-wake patterns from activity recordings

Napping and circadian sleep-wake regulation during healthy aging

STUDY OBJECTIVES

Daytime napping is frequently reported among the older population and has attracted increasing attention due to its association with multiple health conditions. Here, we tested whether napping in the aged is associated with altered circadian regulation of sleep, sleepiness, and vigilance performance.

METHODS

Sixty healthy older individuals (mean age: 69 years, 39 women) were recruited with respect to their napping habits (30 nappers, 30 non-nappers). All participants underwent an in-lab 40-hour multiple nap protocol (10 cycles of 80 minutes of sleep opportunity alternating with 160 minutes of wakefulness), preceded and followed by a baseline and recovery sleep period. Saliva samples for melatonin assessment, sleepiness, and vigilance performance were collected during wakefulness and electrophysiological data were recorded to derive sleep parameters during scheduled sleep opportunities.

RESULTS

The circadian amplitude of melatonin secretion was reduced in nappers, compared to non-nappers. Furthermore, nappers were characterized by higher sleep efficiencies and REM sleep proportion during day- compared to nighttime naps. The nap group also presented altered modulation in sleepiness and vigilance performance at specific circadian phases.

DISCUSSION

Our data indicate that napping is associated with an altered circadian sleep-wake propensity rhythm. They thereby contribute to the understanding of the biological correlates underlying napping and/or sleep-wake cycle fragmentation during healthy aging. Altered circadian sleep-wake promotion can lead to a less distinct allocation of sleep into nighttime and/or a reduced wakefulness drive during the day, thereby potentially triggering the need to sleep at adverse circadian phase.

Validating CircaCP: a generic sleep-wake cycle detection algorithm for unlabelled actigraphy data

Sleep-wake (SW) cycle detection is a key step for extracting temporal sleep metrics from actigraphy. Various supervised learning algorithms have been developed, yet their generalizability from sensor to sensor or study to study is questionable. In this paper, we detail and validate an unsupervised algorithm-CircaCP-for detecting SW cycles from actigraphy. It first uses a robust cosinor model to estimate circadian rhythm, then searches for a single change point (CP) within each circadian cycle. Using CircaCP, we estimated sleep/wake onset times (S/WOTs) from 2125 individuals' data in the MESA sleep study and compared the estimated S/WOTs against self-reported S/WOT event markers, using Bland-Altman analysis as well as variance component analysis. On average, SOTs estimated by CircaCP were 3.6 min behind those reported by event markers, and WOTs by CircaCP were less than 1 min behind those reported by markers. These differences accounted for less than 0.2% variability in S/WOTs, considering other sources of between-subject variations. Rooted in first principles of human circadian rhythms, our algorithm transferred seamlessly from children's hip-worn ActiGraph data to ageing adults' wrist-worn Actiwatch data. The generalizability of our algorithm suggests that it can be widely applied to actigraphy collected by other sensors and studies.

Performance of an open machine learning model to classify sleep/wake from actigraphy across ∼24-hour intervals without knowledge of rest timing

GOAL AND AIMS

Commonly used actigraphy algorithms are designed to operate within a known in-bed interval. However, in free-living scenarios this interval is often unknown. We trained and evaluated a sleep/wake classifier that operates on actigraphy over ∼24-hour intervals, without knowledge of in-bed timing.

FOCUS TECHNOLOGY

Actigraphy counts from ActiWatch Spectrum devices.

REFERENCE TECHNOLOGY

Sleep staging derived from polysomnography, supplemented by observation of wakefulness outside of the staged interval. Classifications from the Oakley actigraphy algorithm were additionally used as performance reference.

SAMPLE

Adults, sleeping in either a home or laboratory environment.

DESIGN

Machine learning was used to train and evaluate a sleep/wake classifier in a supervised learning paradigm. The classifier is a temporal convolutional network, a form of deep neural network.

CORE ANALYTICS

Performance was evaluated across ∼24 hours, and additionally restricted to only in-bed intervals, both in terms of epoch-by-epoch performance, and the discrepancy of summary statistics within the intervals.

ADDITIONAL ANALYTICS AND EXPLORATORY ANALYSES

Performance of the trained model applied to the Multi-Ethnic Study of Atherosclerosis dataset.

CORE OUTCOMES

Over ∼24 hours, the temporal convolutional network classifier produced the same or better performance as the Oakley classifier on all measures tested. When restricting analysis to the in-bed interval, the temporal convolutional network remained favorable on several metrics.

IMPORTANT SUPPLEMENTAL OUTCOMES

Performance decreased on the Multi-Ethnic Study of Atherosclerosis dataset, especially when restricting analysis to the in-bed interval.

CORE CONCLUSION

A classifier using data labeled over ∼24-hour intervals allows for the continuous classification of sleep/wake without knowledge of in-bed intervals. Further development should focus on improving generalization performance.

Estimation of bloodstain deposition time within a 24-h day-night cycle with rhythmic mRNA based on a machine learning algorithm

Estimating the time that bloodstains are left at a crime scene can provide invaluable evidence for law enforcement investigations, including determining the time of the crime, linking the perpetrator to the crime scene, narrowing the pool of possible suspects, and verifying witness statements. There have been some attempts to estimate the time since deposition of bloodstains, i.e., how much time has passed since the bloodstain was left at a crime scene. However, most studies focus on the time interval of days. As far as we know, previous study have been conducted to estimate the deposition time of blood within a 24-h day-night cycle. To date, there is a lack of studies on whether rhythmic mRNA of blood is suitable for bloodstain samples. In this study, we estimated the bloodstain deposition time within a 24-h day-night cycle based on the expression of messenger RNAs (mRNAs) by real-time quantitative polymerase chain reaction. Bloodstain samples were prepared from eight individuals at eight time points under real and uncontrolled conditions. Four mRNAs expressed rhythmically and were used to construct a regression model using the k-nearest neighbor (KNN) algorithm, resulting in a mean absolute error of 3.92 h. Overall, using the rhythmic mRNAs, a machine learning model was developed which has allowed us to predict the deposition time of bloodstains within the 24-h day-night cycle in East Asian populations. This study demonstrates that mRNA biomarkers can be used to estimate the bloodstain deposition time within a 24-h period. Furthermore, rhythmic mRNA biomarkers provide a potential method and perspective for estimating the deposition time of forensic traces in forensic investigation. Case samples in forensic analysis are usually limited or degraded, so the stability and sensitivity of rhythmic biomarkers need to be further investigated.

Food addiction and emotional eating are associated with intradaily rest-activity rhythm variability

PURPOSE

The aim of the present investigation was to study the associations among parameters characterizing eating behavior and actimetry-derived indices of circadian rhythm of motor activity.

METHODS

The study involved 81 healthy participants (average age: 21.5 ± 9.6 y, women: 77.8%). Each study participant provided personal data, filled out the Yale Food Addiction Scale and the Dutch Eating Behavior Questionnaire, and wore a wrist actimeter for 7 consecutive days to record motor activity. Using time series treatments, we obtained: (a) three cosinor-derived parametric indices [Medline Estimating Statistics of Rhythm (MESOR), amplitude, and acrophase], and (b) four non-parametric indices [interdaily stability, intradaily variability (I.V.), most active 10-h period (M10), and least active 5-h period] characterizing the 24-h rhythm of motor activity. A multiple regression analysis adjusted for age, sex, and BMI was performed to assess the associations among the studied indicators.

RESULTS

It was shown that I.V. is a predictor of symptoms of food addiction (β = 0.242, P = 0.037) and emotional eating (β = 0.390, P = 0.004), MESOR is a predictor of symptoms of food addiction (β = 0.342, P = 0.003), and M10 predicts restraint (β = 0.257, P = 0.015) and emotional eating (β = 0.464, P = 0.001).

CONCLUSION

It was shown for the first time that an increase in symptom counts of food addiction is associated with an increase in the average level and fragmentation of 24-h rhythm of motor activity.

LEVEL OF EVIDENCE

Level V, cross-sectional descriptive study.

Wearable Light-and-Motion Dataloggers for Sleep/Wake Research: A Review

Long-term recording of a person’s activity (actimetry or actigraphy) using devices typically worn on the wrist is increasingly applied in sleep/wake, chronobiological, and clinical research to estimate parameters of sleep and sleep-wake cycles. With the recognition of the importance of light in influencing these parameters and with the development of technological capabilities, light sensors have been introduced into devices to correlate physiological and environmental changes. Over the past two decades, many such new devices have appeared from different manufacturers. One of the aims of this review is to help researchers and clinicians choose the data logger that best fits their research goals. Seventeen currently available light-and-motion recorders entered the analysis. They were reviewed for appearance, dimensions, weight, mounting, battery, sensors, features, communication interface, and software. We found that all devices differed from each other in several features. In particular, six devices are equipped with a light sensor that can measure blue light. It is noteworthy that blue light most profoundly influences the physiology and behavior of mammals. As the wearables market is growing rapidly, this review helps guide future developments and needs to be updated every few years.

Smartphone keyboard interaction monitoring as an unobtrusive method to approximate rest-activity patterns: Inter-individual and metric-specific variations (Preprint)

BACKGROUND

Sleep is an important determinant of individuals' health and behavior during the wake phase. Novel research methods for field assessments are required to enable the monitoring of sleep over a prolonged period and across a large number of people. The ubiquity of smartphones offers new avenues for detecting rest-activity patterns in everyday life in a noninvasive an inexpensive manner and on a large scale. Recent studies provided evidence for the potential of smartphone interaction monitoring as a novel tracking method to approximate rest-activity patterns based on the timing of smartphone activity and inactivity throughout the 24-hour day. These findings require further replication and more detailed insights into interindividual variations in the associations and deviations with commonly used metrics for monitoring rest-activity patterns in everyday life.

OBJECTIVE

This study aimed to replicate and expand on earlier findings regarding the associations and deviations between smartphone keyboard-derived and self-reported estimates of the timing of the onset of the rest and active periods and the duration of the rest period. Moreover, we aimed to quantify interindividual variations in the associations and time differences between the 2 assessment modalities and to investigate to what extent general sleep quality, chronotype, and trait self-control moderate these associations and deviations.

METHODS

Students were recruited to participate in a 7-day experience sampling study with parallel smartphone keyboard interaction monitoring. Multilevel modeling was used to analyze the data.

RESULTS

In total, 157 students participated in the study, with an overall response rate of 88.9% for the diaries. The results revealed moderate to strong relationships between the keyboard-derived and self-reported estimates, with stronger associations for the timing-related estimates (β ranging from .61 to .78) than for the duration-related estimates (β=.51 and β=.52). The relational strength between the time-related estimates was lower, but did not substantially differ for the duration-related estimates, among students experiencing more disturbances in their general sleep quality. Time differences between the keyboard-derived and self-reported estimates were, on average, small (<0.5 hours); however, large discrepancies were also registered for quite some nights. The time differences between the 2 assessment modalities were larger for both timing-related and rest duration-related estimates among students who reported more disturbances in their general sleep quality. Chronotype and trait self-control did not significantly moderate the associations and deviations between the 2 assessment modalities.

CONCLUSIONS

We replicated the positive potential of smartphone keyboard interaction monitoring for estimating rest-activity patterns among populations of regular smartphone users. Chronotype and trait self-control did not significantly influence the metrics' accuracy, whereas general sleep quality did: the behavioral proxies obtained from smartphone interactions appeared to be less powerful among students who experienced lower general sleep quality. The generalization and underlying process of these findings require further investigation.

pyActigraphy: Open-source python package for actigraphy data visualization and analysis

Over the past 40 years, actigraphy has been used to study rest-activity patterns in circadian rhythm and sleep research. Furthermore, considering its simplicity of use, there is a growing interest in the analysis of large population-based samples, using actigraphy. Here, we introduce pyActigraphy, a comprehensive toolbox for data visualization and analysis including multiple sleep detection algorithms and rest-activity rhythm variables. This open-source python package implements methods to read multiple data formats, quantify various properties of rest-activity rhythms, visualize sleep agendas, automatically detect rest periods and perform more advanced signal processing analyses. The development of this package aims to pave the way towards the establishment of a comprehensive open-source software suite, supported by a community of both developers and researchers, that would provide all the necessary tools for in-depth and large scale actigraphy data analyses.

The possibility to continuously record locomotor movements using accelerometers (actigraphy) has allowed field studies of sleep and rest-activity patterns. It has also enabled large-scale data collections, opening new avenues for research. However, each brand of actigraph devices encodes recordings in its own format and closed-source proprietary softwares are typically used to read and analyse actigraphy data. In order to provide an alternative to these softwares, we developed a comprehensive open-source toolbox for actigraphy data analysis, pyActigraphy. It allows researchers to read actigraphy data from 7 different file formats and gives access to a variety of rest-activity rhythm variables, automatic sleep detection algorithms and more advanced signal processing techniques. Besides, in order to empower researchers and clinicians with respect to their analyses, we created a series of interactive tutorials that illustrate how to implement the key steps of typical actigraphy data analyses. As an open-source project, all kind of user’s contributions to our toolbox are welcome. As increasing evidence points to the predicting value of rest-activity patterns derived from actigraphy for brain integrity, we believe that the development of the pyActigraphy package will not only benefit the sleep and chronobiology research, but also the neuroscientific community at large.

Wrist actigraphic approach in primary, secondary and tertiary care based on the principles of predictive, preventive and personalised (3P) medicine

Abstract

Sleep quality and duration as well as activity-rest-cycles at individual level are crucial for maintaining physical and mental health. Although several methods do exist to monitor these parameters, optimal approaches are still under consideration and technological development. Wrist actigraphy is a non-invasive electro-physical method validated in the field of chronobiology to record movements and to allow for monitoring human activity-rest-cycles. Based on the continuous recording of motor activity and light exposure, actigraphy provides valuable information about the quality and quantity of the sleep–wake rhythm and about the amount of motor activity at day and night that is highly relevant for predicting a potential disease and its targeted prevention as well as personalisation of medical services provided to individuals in suboptimal health conditions and patients. Being generally used in the field of sleep medicine, actigraphy demonstrates a great potential to be successfully implemented in primary, secondary and tertiary care, psychiatry, oncology, and intensive care, military and sports medicines as well as epidemiological monitoring of behavioural habits as well as well-being medical support, amongst others.

Prediction of disease development and individual outcomes

Activity-rest-cycles have been demonstrated to be an important predictor for many diseases including but not restricted to the development of metabolic, psychiatric and malignant pathologies. Moreover, activity-rest-cycles directly impact individual outcomes in corresponding patient cohorts.

Targeted prevention

Data acquired by actigraphy are instrumental for the evidence-based targeted prevention by analysing individualised patient profiles including light exposure, sleep duration and quality, activity-rest-cycles, intensity and structure of motion pattern.

Personalised therapy

Wrist actigraphic approach is increasingly used in clinical care. Personalised measurements of sedation/agitation rhythms are useful for ICU patients, for evaluation of motor fatigue in oncologic patients, for an individual enhancement of performance in military and sport medicine. In the framework of personalised therapy intervention, patients can be encouraged to optimise their behavioural habits improving recovery and activity patterns. This opens excellent perspectives for the sleep-inducing medication and stimulants replacement as well as for increasing the role of participatory medicine by visualising and encouraging optimal behavioural patterns of the individual.