Hyoki Lee

Improving Sleep Quality Assessment Using Wearable Sensors by Including Information From Postural/Sleep Position Changes and Body Acceleration: A Comparison of Chest-Worn Sensors, Wrist Actigraphy, and Polysomnography

STUDY OBJECTIVES To improve sleep quality assessment using a single chest-worn sensor by extracting body acceleration and sleep position changes. METHODS Sleep patterns of 21 participants (50.8 ± 12.8 years, 47.8% female) with self-reported sleep problems were simultaneously recorded using a chest sensor (Chest), an Actiwatch (Wrist), and polysomnography (PSG) during overnight sleep laboratory assessment. An algorithm for Chest was developed to detect sleep/wake epochs based on body acceleration and sleep position/postural changes data, which were then used to estimate sleep parameters of interest. Comparisons between Chest and Wrist with respect to PSG were performed. Identification of sleep/wake epochs was assessed by estimating sensitivity, specificity, and accuracy. Agreement between sensor-derived sleep parameters and PSG was assessed using correlation coefficients and Bland-Altman analysis. RESULTS Chest identified sleep/wake epochs with an accuracy of on average 6% higher than Wrist (85.8% versus 79.8%). Similar trends were observed for sensitivity/specificity values. Correlation between Wrist and PSG was poor for most of the sleep parameters of interest (r = 0.0-0.3); however, Chest and PSG correlation showed moderate to strong agreement (r = 0.4-0.8) with relatively low bias and high precision bias (precision): 9.2 (13.2) minutes for sleep onset latency; 17.3(34.8) minutes for total sleep time; 7.5 (29.8) minutes for wake after sleep onset; and 2.0 (7.3)% for sleep efficacy. CONCLUSIONS Combination of sleep postural/position changes and body acceleration improved detection of sleep/wake epochs compared to wrist acceleration alone. The chest sensors also improved estimation of sleep parameters of interest with stronger agreement with PSG. Our findings may expand the application of wearable sensors to clinically assess sleep outside of a sleep laboratory.

Motor Planning Error: Toward Measuring Cognitive Frailty in Older Adults Using Wearables

Practical tools which can be quickly administered are needed for measuring subtle changes in cognitive–motor performance over time. Frailty together with cognitive impairment, or ‘cognitive frailty’, are shown to be strong and independent predictors of cognitive decline over time. We have developed an interactive instrumented trail-making task (iTMT) platform, which allows quantification of motor planning error (MPE) through a series of ankle reaching tasks. In this study, we examined the accuracy of MPE in identifying cognitive frailty in older adults. Thirty-two older adults (age = 77.3 ± 9.1 years, body-mass-index = 25.3 ± 4.7 kg/m2, female = 38%) were recruited. Using either the Mini-Mental State Examination or Montreal Cognitive Assessment (MoCA), 16 subjects were classified as cognitive-intact and 16 were classified as cognitive-impaired. In addition, 12 young-healthy subjects (age = 26.0 ± 5.2 years, body-mass-index = 25.3 ± 3.9 kg/m2, female = 33%) were recruited to establish a healthy benchmark. Subjects completed the iTMT, using an ankle-worn sensor, which transforms ankle motion into navigation of a computer cursor. The iTMT task included reaching five indexed target circles (including numbers 1-to-3 and letters A&B placed in random order) on the computer-screen by moving the ankle-joint while standing. The ankle-sensor quantifies MPE through analysis of the pattern of ankle velocity. MPE was defined as percentage of time deviation between subject’s maximum ankle velocity and the optimal maximum ankle velocity, which is halfway through the reaching pathway. Data from gait tests, including single task and dual task walking, were also collected to determine cognitive–motor performance. The average MPE in young-healthy, elderly cognitive-intact, and elderly cognitive-impaired groups was 11.1 ± 5.7%, 20.3 ± 9.6%, and 34.1 ± 4.2% (p < 0.001), respectively. Large effect sizes (Cohen’s d = 1.17–4.56) were observed for discriminating between groups using MPE. Significant correlations were observed between the MPE and MoCA score (r = −0.670, p < 0.001) as well as between the MPE and dual task stride velocity (r = −0.584, p < 0.001). This study demonstrated feasibility and efficacy of estimating MPE from a practical wearable platform with promising results in identifying cognitive–motor impairment and potential application in assessing cognitive frailty. The proposed platform could be also used as an alternative to dual task walking test, where gait assessment may not be practical. Future studies need to confirm these observations in larger samples.

Toward Using a Smartwatch to Monitor Frailty in a Hospital Setting: Using a Single Wrist-Wearable Sensor to Assess Frailty in Bedbound Inpatients

Background: While various objective tools have been validated for assessing physical frailty in the geriatric population, these are often unsuitable for busy clinics and mobility-impaired patients. Recently, we have developed a frailty meter (FM) using two wearable sensors, which allows capturing key frailty phenotypes (weakness, slowness, and exhaustion), by testing 20-s rapid elbow flexion-extension test. Objective: In this study, we proposed an enhanced automated algorithm to identify frailty using a single wrist-worn sensor. Methods: The data collected from 100 geriatric inpatients (age: 78.9 ± 9.1 years, 49% frail) were reanalyzed to validate the new algorithm. The frailty status of the participants was determined using a validated modified frailty index. Different FM phenotypes (31 features) including velocity of elbow rotation, decline in velocity of elbow rotation over 20 s, range of motion, etc. were extracted. A regression model, bootstrap with 2,000 iterations, and recursive feature elimination technique were used for optimizing the FM parameters and identifying frailty using a single wrist-worn sensor. Results: A strong agreement was observed between two-sensor and wrist-worn sensor configuration (r = 0.87, p < 0.001). Results suggest that the wrist-worn FM with no demographic information still yields a high accuracy of 80.0% (95% CI: 79.7-80.3%) and an area under the curve of 87.7% (95% CI: 87.4-87.9%) to identify frailty status. Results are comparable with two-sensor configuration, where the observed accuracy and area under the curve were 80.6% (95% CI: 80.4-80.9%) and 87.4% (95% CI: 87.1-87.6%), respectively. Conclusion: The simplicity of FM may open new avenues to integrate wearable technology and mobile health to capture frailty status in a busy hospital setting. Furthermore, the reduction of needed sensors to a single wrist-worn sensor allows deployment of the proposed algorithm in the form of a smartwatch application. From the application standpoint, the proposed FM is superior to traditional physical frailty-screening tools in which the walking test is a key frailty phenotype, and thus they cannot be used for bedbound patients or in busy clinics where administration of gait test as a part of routine assessment is impractical.

Feature Importance and Predictive Modeling for Multi-source Healthcare Data with Missing Values

With rapid development of sensor technologies and the internet of things, research in the area of connected health is increasing in importance and complexity with wide-reaching impacts for public health. As data sources such as mobile (wearable) sensors get cheaper, smaller, and smarter, important research questions can be answered by combining information from multiple data sources. However, integration of multiple heterogeneous data streams often results in a dataset with several empty cells or missing values. The challenge is to use such sparsely populated integrated datasets without compromising model performance. Naïve approaches for dataset modification such as discarding observations or ad-hoc replacement of missing values often lead to misleading results. In this paper, we discuss and evaluate current best-practices for modeling such data with missing values and then propose an ensemble-learning based sparse-data modeling framework. We develop a predictive model using this framework and compare it with existing models using a study in a healthcare setting. Instead of generating a single score on variable/feature importance, our framework enables the user to understand the importance of a variable based on the existing data values and their localized impact on the outcome.

Automatic classification of apnea/hypopnea events through sleep/wake states and severity of SDB from a pulse oximeter.

This study proposes a method of automatically classifying sleep apnea/hypopnea events based on sleep states and the severity of sleep-disordered breathing (SDB) using photoplethysmogram (PPG) and oxygen saturation (SpO2) signals acquired from a pulse oximeter. The PPG was used to classify sleep state, while the severity of SDB was estimated by detecting events of SpO2 oxygen desaturation. Furthermore, we classified sleep apnea/hypopnea events by applying different categorisations according to the severity of SDB based on a support vector machine. The classification results showed sensitivity performances and positivity predictive values of 74.2% and 87.5% for apnea, 87.5% and 63.4% for hypopnea, and 92.4% and 92.8% for apnea + hypopnea, respectively. These results represent better or comparable outcomes compared to those of previous studies. In addition, our classification method reliably detected sleep apnea/hypopnea events in all patient groups without bias in particular patient groups when our algorithm was applied to a variety of patient groups. Therefore, this method has the potential to diagnose SDB more reliably and conveniently using a pulse oximeter.

Sleep Apnea Detection using Estimated Stroke Volume

Abstract: This paper proposes a new algorithm for sleep apnea detection based on stroke volume. It is very impor-tant to detect sleep apnea since it is a common and serious sleep-disordered breathing (SDB). In the previous studies,methods for sleep apnea detection using heart rate variability, airflow and blood oxygen saturation, tracheal soundhave been proposed, but a method using stroke volume has not been studied. The proposed algorithm consists ofdetection of characteristic points in continuous blood pressure signal, estimation of stroke volume and detection ofsleep apnea. To evaluate the performance of algorithm, the MIT-BIH Polysomnographic Database provided by Phsio-Net was used. As a result, the sensitivity of 85.99%, the specificity of 72.69%, and the accuracy of 84.34%, on theaverage were obtained. The proposed method showed comparable or higher performance compared with previousmethods.Key words: Sleep apnea, Stroke volume, Hypopnea, Obstructive sleep apnea (OSA), Central sleep apnea (CSA)

Toward smart footwear to track frailty phenotypes—using propulsion performance to determine frailty

Frailty assessment is dependent on the availability of trained personnel and it is currently limited to clinic and supervised setting. The growing aging population has made it necessary to find phenotypes of frailty that can be measured in an unsupervised setting for translational application in continuous, remote, and in-place monitoring during daily living activity, such as walking. We analyzed gait performance of 161 older adults using a shin-worn inertial sensor to investigate the feasibility of developing a foot-worn sensor to assess frailty. Sensor-derived gait parameters were extracted and modeled to distinguish different frailty stages, including non-frail, pre-frail, and frail, as determined by Fried Criteria. An artificial neural network model was implemented to evaluate the accuracy of an algorithm using a proposed set of gait parameters in predicting frailty stages. Changes in discriminating power was compared between sensor data extracted from the left and right shin sensor. The aim was to investigate the feasibility of developing a foot-worn sensor to assess frailty. The results yielded a highly accurate model in predicting frailty stages, irrespective of sensor location. The independent predictors of frailty stages were propulsion duration and acceleration, heel-off and toe-off speed, mid stance and mid swing speed, and speed norm. The proposed model enables discriminating different frailty stages with area under curve ranging between 83.2–95.8%. Furthermore, results from the neural network suggest the potential of developing a single-shin worn sensor that would be ideal for unsupervised application and footwear integration for continuous monitoring during walking.

Estimation of Respiration Using Photoplethysmograph During Sleep

Respiratory signal is one of the important physiological information indicating the status and function of the body. Recent studies have provided the possibility of being able to estimate the respiratory signal by using a change of PWV(pulse width variability), PRV(pulse rate variability) and PAV(pulse amplitude variability) in the PPG (photoplethysmography) signal during daily life. But, it is not clear whether the respiratory monitoring is possible even during sleep. Therefore, in this paper, we estimated the respiration from PWV, PRV and PAV of PPG signals during sleep. In addition, respiration rates of the estimated respiration signal were calculated through a time-fre- quency analysis, and errors between respiration rates calculated from each parameter and from reference signal were evaluated in terms of 1 sec, 10 sec and 1 min. As a result, it showed the errors in PWV(1s: 36.38 ± 37.69 mHz, 10s: 36.53 ± 38.16 mHz, 60s: 30.35 ± 38.72 mHz), in PRV(1s: 1.45 ± 1.38 mHz, 10s: 1.44 ± 1.37 mHz, 60s: 0.45 ± 0.56 mHz), and in PAV(1s: 1.05 ± 0.81 mHz, 10s: 1.05 ± 0.79 mHz, 60s: 0.56 ± 0.93 mHz). The errors in PRV and PAV are lower than that of PWV. Finally, we concluded that PRV and PAV are more effective than PWV in monitoring the respiration in daily life as well as during sleep.

Effects of office workstation type on physical activity and stress

Objective Office environments have been causally linked to workplace-related illnesses and stress, yet little is known about how office workstation type is linked to objective metrics of physical activity and stress. We aimed to explore these associations among office workers in US federal office buildings. Methods We conducted a wearable, sensor-based, observational study of 231 workers in four office buildings. Outcome variables included workers’ physiological stress response, physical activity and perceived stress. Relationships between office workstation type and these variables were assessed using structural equation modelling. Results Workers in open bench seating were more active at the office than those in private offices and cubicles (open bench seating vs private office=225.52 mG (31.83% higher on average) (95% CI 136.57 to 314.46); open bench seating vs cubicle=185.13 mG (20.16% higher on average) (95% CI 66.53 to 303.72)). Furthermore, workers in open bench seating experienced lower perceived stress at the office than those in cubicles (−0.27 (9.10% lower on average) (95% CI −0.54 to −0.02)). Finally, higher physical activity at the office was related to lower physiological stress (higher heart rate variability in the time domain) outside the office (−26.12 ms/mG (14.18% higher on average) (95% CI −40.48 to −4.16)). Conclusions Office workstation type was related to enhanced physical activity and reduced physiological and perceived stress. This research highlights how office design, driven by office workstation type, could be a health-promoting factor.

A Regularization Approach for Identifying Cumulative Lagged Effects in Smart Health Applications

Recent development of wearable sensor technologies have made it possible to capture concurrent data streams for ambient environment and instantaneous physiological stress response at a fine granularity. Characterizing the delay in physiological stress response time to each environment stimulus is as important as capturing the magnitude of the effect. In this paper, we discuss and evaluate a new regularization-based statistical method to determine the ideal lagged effect of five environmental factors-carbon dioxide, temperature, relative humidity, atmospheric pressure and noise levels on instantaneous stress response. Using this method, we infer that the first four environment variables have a cumulative lagged effect, of approximately 60 minutes, on stress response whereas noise level has an instantaneous effect on stress response. The proposed transformations to inputs result in models with better fit and predictive performance. This study not only informs the field of environment-wellbeing research about the cumulative lagged effects of the specified environmental factors, but also proposes a new method for determining optimal feature transformation in similar smart health studies.