UMAHand: A dataset of inertial signals of typical hand activities.

Given the popularity of wrist-worn devices, particularly smartwatches, the identification of manual movement patterns has become of utmost interest within the research field of Human Activity Recognition (HAR) systems. In this context, by leveraging the numerous sensors natively embedded in smartwatches, the HAR functionalities that can be implemented in a watch via software and in a very cost-efficient way cover a wide variety of applications, ranging from fitness trackers to gesture detectors aimed at disabled individuals (e.g., for sending alarms), promoting behavioral activation or healthy lifestyle habits. In this regard, for the development of artificial intelligence algorithms capable of effectively discriminating these activities, it is of great importance to have repositories of movements that allow the scientific community to train, evaluate, and benchmark new proposals of movement detectors. The UMAHand dataset offers a collection of files containing the signals captured by a Shimmer 3 sensor node, which includes an accelerometer, a gyroscope, a magnetometer and a barometer, during the execution of different typical hand movements. For that purpose, the measurements from these four sensors, gathered at a sampling rate of 100 Hz, were taken from a group of 25 volunteers (16 females and 9 males), aged between 18 and 56, during the performance of 29 daily life activities involving hand mobility. Participants wore the sensor node on their dominant hand throughout the experiments.

Estimation of Heart Rate and Energy Expenditure Using a Smart Bracelet during Different Exercise Intensities: A Reliability and Validity Study.

Background. With wrist-worn wearables becoming increasingly available, it is important to understand their reliability and validity in different conditions. The primary objective of this study was to examine the reliability and validity of the Lexin Mio smart bracelet in measuring heart rate (HR) and energy expenditure (EE) in people with different physical activity levels exercising at different intensities. Methods. A total of 65 participants completed one maximal oxygen uptake test and two running exercise tests wearing the Mio smart bracelet, the Polar H10 HR band, and a gas-analysis system. Results. In terms of HR measurement reliability, the Mio smart bracelet showed good reliability in a left versus right test and good test−retest reliability (p > 0.05; mean absolute percentage error (MAPE) < 10%; intraclass correlation coefficient (ICC) > 0.4). For EE measurement, the Mio smart bracelet showed good reliability in a left versus right test, good test−retest reliability on the right (p > 0.05; MAPE > 10%; ICC > 0.4), and low test−retest reliability on the left (p > 0.05; MAPE > 10%; ICC < 0.4). Regarding validity, the Mio smart bracelet showed good validity for HR measurement (p > 0.05; MAPE < 10%; ICC > 0.4) and low validity for EE measurement (p < 0.05; MAPE > 10%; ICC < 0.4). Conclusion. The Lexin Mio smart bracelet showed good reliability and validity for HR measurement among people with different physical activity levels exercising at various exercise intensities in a laboratory setting. However, the smart bracelet showed good reliability and low validity for the estimation of EE.

Transfer learning from ECG to PPG for improved sleep staging from wrist-worn wearables.

Objective.To develop a sleep staging method from wrist-worn accelerometry and the photoplethysmogram (PPG) by leveraging transfer learning from a large electrocardiogram (ECG) database.Approach.In previous work, we developed a deep convolutional neural network for sleep staging from ECG using the cross-spectrogram of ECG-derived respiration and instantaneous beat intervals, heart rate variability metrics, spectral characteristics, and signal quality measures derived from 5793 subjects in Sleep Heart Health Study (SHHS). We updated the weights of this model by transfer learning using PPG data derived from the Empatica E4 wristwatch worn by 105 subjects in the 'Emory Twin Study Follow-up' (ETSF) database, for whom overnight polysomnographic (PSG) scoring was available. The relative performance of PPG, and actigraphy (Act), plus combinations of these two signals, with and without transfer learning was assessed.Main results.The performance of our model with transfer learning showed higher accuracy (1-9 percentage points) and Cohen's Kappa (0.01-0.13) than those without transfer learning for every classification category. Statistically significant, though relatively small, incremental differences in accuracy occurred for every classification category as tested with the McNemar test. The out-of-sample classification performance using features from PPG and actigraphy for four-class classification was Accuracy (Acc) = 68.62% and Kappa = 0.44. For two-class classification, the performance was Acc = 81.49% and Kappa = 0.58.Significance.We proposed a combined PPG and actigraphy-based sleep stage classification approach using transfer learning from a large ECG sleep database. Results demonstrate that the transfer learning approach improves estimates of sleep state. The use of automated beat detectors and quality metrics means human over-reading is not required, and the approach can be scaled for large cross-sectional or longitudinal studies using wrist-worn devices for sleep staging.

Using Smart Bracelets to Assess Heart Rate Among Students During Physical Education Lessons: Feasibility, Reliability, and Validity Study.

BACKGROUND: An increasing number of wrist-worn wearables are being examined in the context of health care. However, studies of their use during physical education (PE) lessons remain scarce. OBJECTIVE: We aim to examine the reliability and validity of the Fizzo Smart Bracelet (Fizzo) in measuring heart rate (HR) in the laboratory and during PE lessons. METHODS: In Study 1, 11 healthy subjects (median age 22.0 years, IQR 3.75 years) twice completed a test that involved running on a treadmill at 6 km/h for 12 minutes and 12 km/h for 5 minutes. During the test, participants wore two Fizzo devices, one each on their left and right wrists, to measure their HR. At the same time, the Polar Team2 Pro (Polar), which is worn on the chest, was used as the standard. In Study 2, we went to 10 schools and measured the HR of 24 students (median age 14.0 years, IQR 2.0 years) during PE lessons. During the PE lessons, each student wore a Polar device on their chest and a Fizzo on their right wrist to measure HR data. At the end of the PE lessons, the students and their teachers completed a questionnaire where they assessed the feasibility of Fizzo. The measurements taken by the left wrist Fizzo and the right wrist Fizzo were compared to estimate reliability, while the Fizzo measurements were compared to the Polar measurements to estimate validity. To measure reliability, intraclass correlation coefficients (ICC), mean difference (MD), standard error of measurement (SEM), and mean absolute percentage errors (MAPE) were used. To measure validity, ICC, limits of agreement (LOA), and MAPE were calculated and Bland-Altman plots were constructed. Percentage values were used to estimate the feasibility of Fizzo. RESULTS: The Fizzo showed excellent reliability and validity in the laboratory and moderate validity in a PE lesson setting. In Study 1, reliability was excellent (ICC>0.97; MD<0.7; SEM<0.56; MAPE<1.45%). The validity as determined by comparing the left wrist Fizzo and right wrist Fizzo was excellent (ICC>0.98; MAPE<1.85%). Bland-Altman plots showed a strong correlation between left wrist Fizzo measurements (bias=0.48, LOA=-3.94 to 4.89 beats per minute) and right wrist Fizzo measurements (bias=0.56, LOA=-4.60 to 5.72 beats per minute). In Study 2, the validity of the Fizzo was lower compared to that found in Study 1 but still moderate (ICC>0.70; MAPE<9.0%). The Fizzo showed broader LOA in the Bland-Altman plots during the PE lessons (bias=-2.60, LOA=-38.89 to 33.69 beats per minute). Most participants considered the Fizzo very comfortable and easy to put on. All teachers thought the Fizzo was helpful. CONCLUSIONS: When participants ran on a treadmill in the laboratory, both left and right wrist Fizzo measurements were accurate. The validity of the Fizzo was lower in PE lessons but still reached a moderate level. The Fizzo is feasible for use during PE lessons.

Use of Free-Living Step Count Monitoring for Heart Failure Functional Classification: Validation Study.

BACKGROUND: The New York Heart Association (NYHA) functional classification system has poor inter-rater reproducibility. A previously published pilot study showed a statistically significant difference between the daily step counts of heart failure (with reduced ejection fraction) patients classified as NYHA functional class II and III as measured by wrist-worn activity monitors. However, the study's small sample size severely limits scientific confidence in the generalizability of this finding to a larger heart failure (HF) population. OBJECTIVE: This study aimed to validate the pilot study on a larger sample of patients with HF with reduced ejection fraction (HFrEF) and attempt to characterize the step count distribution to gain insight into a more objective method of assessing NYHA functional class. METHODS: We repeated the analysis performed during the pilot study on an independently recorded dataset comprising a total of 50 patients with HFrEF (35 NYHA II and 15 NYHA III) patients. Participants were monitored for step count with a Fitbit Flex for a period of 2 weeks in a free-living environment. RESULTS: Comparing group medians, patients exhibiting NYHA class III symptoms had significantly lower recorded 2-week mean daily total step count (3541 vs 5729 [steps], P=.04), lower 2-week maximum daily total step count (10,792 vs 5904 [steps], P=.03), lower 2-week recorded mean daily mean step count (4.0 vs 2.5 [steps/minute], P=.04,), and lower 2-week mean and 2-week maximum daily per minute step count maximums (88.1 vs 96.1 and 111.0 vs 123.0 [steps/minute]; P=.02 and .004, respectively). CONCLUSIONS: Patients with NYHA II and III symptoms differed significantly by various aggregate measures of free-living step count including the (1) mean and (2) maximum daily total step count as well as by the (3) mean of daily mean step count and by the (4) mean and (5) maximum of the daily per minute step count maximum. These findings affirm that the degree of exercise intolerance of NYHA II and III patients as a group is quantifiable in a replicable manner. This is a novel and promising finding that suggests the existence of a possible, completely objective measure of assessing HF functional class, something which would be a great boon in the continuing quest to improve patient outcomes for this burdensome and costly disease.

About the Accuracy and Problems of Consumer Devices in the Assessment of Sleep.

Commercial sleep devices and mobile-phone applications for scoring sleep are gaining ground. In order to provide reliable information about the quantity and/or quality of sleep, their performance needs to be assessed against the current gold standard, i.e., polysomnography (PSG; measuring brain, eye, and muscle activity). Here, we assessed some commercially available sleep trackers, namely an activity tracker; Mi band (Xiaomi, Beijing, China), a scientific actigraph: Motionwatch 8 (CamNTech, Cambridge, UK), and a much-used mobile phone application: Sleep Cycle (Northcube, Gothenburg, Sweden). We recorded 27 nights in healthy sleepers using PSG and these devices and compared the results. Surprisingly, all devices had poor agreement with the PSG gold standard. Sleep parameter comparisons revealed that, specifically, the Mi band and the Sleep Cycle application had difficulties in detecting wake periods which negatively affected their total sleep time and sleep-efficiency estimations. However, all 3 devices were good in detecting the most basic parameter, the actual time in bed. In summary, our results suggest that, to date, the available sleep trackers do not provide meaningful sleep analysis but may be interesting for simply tracking time in bed. A much closer interaction with the scientific field seems necessary if reliable information shall be derived from such devices in the future.

Accuracy of Apple Watch Measurements for Heart Rate and Energy Expenditure in Patients With Cardiovascular Disease: Cross-Sectional Study.

BACKGROUND: Wrist-worn tracking devices such as the Apple Watch are becoming more integrated in health care. However, validation studies of these consumer devices remain scarce. OBJECTIVES: This study aimed to assess if mobile health technology can be used for monitoring home-based exercise in future cardiac rehabilitation programs. The purpose was to determine the accuracy of the Apple Watch in measuring heart rate (HR) and estimating energy expenditure (EE) during a cardiopulmonary exercise test (CPET) in patients with cardiovascular disease. METHODS: Forty patients (mean age 61.9 [SD 15.2] yrs, 80% male) with cardiovascular disease (70% ischemic, 22.5% valvular, 7.5% other) completed a graded maximal CPET on a cycle ergometer while wearing an Apple Watch. A 12-lead electrocardiogram (ECG) was used to measure HR; indirect calorimetry was used for EE. HR was analyzed at three levels of intensity (seated rest, HR1; moderate intensity, HR2; maximal performance, HR3) for 30 seconds. The EE of the entire test was used. Bias or mean difference (MD), standard deviation of difference (SDD), limits of agreement (LoA), mean absolute error (MAE), mean absolute percentage error (MAPE), and intraclass correlation coefficients (ICCs) were calculated. Bland-Altman plots and scatterplots were constructed. RESULTS: SDD for HR1, HR2, and HR3 was 12.4, 16.2, and 12.0 bpm, respectively. Bias and LoA (lower, upper LoA) were 3.61 (-20.74, 27.96) for HR1, 0.91 (-30.82, 32.63) for HR2, and -1.82 (-25.27, 21.63) for HR3. MAE was 6.34 for HR1, 7.55 for HR2, and 6.90 for HR3. MAPE was 10.69% for HR1, 9.20% for HR2, and 6.33% for HR3. ICC was 0.729 (P<.001) for HR1, 0.828 (P<.001) for HR2, and 0.958 (P<.001) for HR3. Bland-Altman plots and scatterplots showed good correlation without systematic error when comparing Apple Watch with ECG measurements. SDD for EE was 17.5 kcal. Bias and LoA were 30.47 (-3.80, 64.74). MAE was 30.77; MAPE was 114.72%. ICC for EE was 0.797 (P<.001). The Bland-Altman plot and a scatterplot directly comparing Apple Watch and indirect calorimetry showed systematic bias with an overestimation of EE by the Apple Watch. CONCLUSIONS: In patients with cardiovascular disease, the Apple Watch measures HR with clinically acceptable accuracy during exercise. If confirmed, it might be considered safe to incorporate the Apple Watch in HR-guided training programs in the setting of cardiac rehabilitation. At this moment, however, it is too early to recommend the Apple Watch for cardiac rehabilitation. Also, the Apple Watch systematically overestimates EE in this group of patients. Caution might therefore be warranted when using the Apple Watch for measuring EE.

Assessing the Use of Wrist-Worn Devices in Patients With Heart Failure: Feasibility Study.

BACKGROUND: Exercise capacity and raised heart rate (HR) are important prognostic markers in patients with heart failure (HF). There has been significant interest in wrist-worn devices that track activity and HR. OBJECTIVE: We aimed to assess the feasibility and accuracy of HR and activity tracking of the Fitbit and Apple Watch. METHODS: We conducted a two-phase study assessing the accuracy of HR by Apple Watch and Fitbit in healthy participants. In Phase 1, 10 healthy individuals wore a Fitbit, an Apple Watch, and a GE SEER Light 5-electrode Holter monitor while exercising on a cycle ergometer with a 10-watt step ramp protocol from 0-100 watts. In Phase 2, 10 patients with HF and New York Heart Association (NYHA) Class II-III symptoms wore wrist devices for 14 days to capture overall step count/exercise levels. RESULTS: Recorded HR by both wrist-worn devices had the best agreement with Holter readings at a workload of 60-100 watts when the rate of change of HR is less dynamic. Fitbit recorded a mean 8866 steps/day for NYHA II patients versus 4845 steps/day for NYHA III patients (P=.04). In contrast, Apple Watch recorded a mean 7027 steps/day for NYHA II patients and 4187 steps/day for NYHA III patients (P=.08). CONCLUSIONS: Both wrist-based devices are best suited for static HR rate measurements. In an outpatient setting, these devices may be adequate for average HR in patients with HF. When assessing exercise capacity, the Fitbit better differentiated patients with NYHA II versus NYHA III by the total number of steps recorded. This exploratory study indicates that these wrist-worn devices show promise in prognostication of HF in the continuous monitoring of outpatients.