Early adverse physiological event detection using commercial wearables: challenges and opportunities.

Data from commercial off-the-shelf (COTS) wearables leveraged with machine learning algorithms provide an unprecedented potential for the early detection of adverse physiological events. However, several challenges inhibit this potential, including (1) heterogeneity among and within participants that make scaling detection algorithms to a general population less precise, (2) confounders that lead to incorrect assumptions regarding a participant's healthy state, (3) noise in the data at the sensor level that limits the sensitivity of detection algorithms, and (4) imprecision in self-reported labels that misrepresent the true data values associated with a given physiological event. The goal of this study was two-fold: (1) to characterize the performance of such algorithms in the presence of these challenges and provide insights to researchers on limitations and opportunities, and (2) to subsequently devise algorithms to address each challenge and offer insights on future opportunities for advancement. Our proposed algorithms include techniques that build on determining suitable baselines for each participant to capture important physiological changes and label correction techniques as it pertains to participant-reported identifiers. Our work is validated on potentially one of the largest datasets available, obtained with 8000+ participants and 1.3+ million hours of wearable data captured from Oura smart rings. Leveraging this extensive dataset, we achieve pre-symptomatic detection of COVID-19 with a performance receiver operator characteristic (ROC) area under the curve (AUC) of 0.725 without correction techniques, 0.739 with baseline correction, 0.740 with baseline correction and label correction on the training set, and 0.777 with baseline correction and label correction on both the training and the test set. Using the same respective paradigms, we achieve ROC AUCs of 0.919, 0.938, 0.943 and 0.994 for the detection of self-reported fever, and 0.574, 0.611, 0.601, and 0.635 for detection of self-reported shortness of breath. These techniques offer improvements across almost all metrics and events, including PR AUC, sensitivity at 75% specificity, and precision at 75% recall. The ring allows continuous monitoring for detection of event onset, and we further demonstrate an improvement in the early detection of COVID-19 from an average of 3.5 days to an average of 4.1 days before a reported positive test result.

Physics-informed neural networks for modeling physiological time series for cuffless blood pressure estimation.

The bold vision of AI-driven pervasive physiological monitoring, through the proliferation of off-the-shelf wearables that began a decade ago, has created immense opportunities to extract actionable information for precision medicine. These AI algorithms model input-output relationships of a system that, in many cases, exhibits complex nature and personalization requirements. A particular example is cuffless blood pressure estimation using wearable bioimpedance. However, these algorithms need training over significant amount of ground truth data. In the context of biomedical applications, collecting ground truth data, particularly at the personalized level is challenging, burdensome, and in some cases infeasible. Our objective is to establish physics-informed neural network (PINN) models for physiological time series data that would use minimal ground truth information to extract complex cardiovascular information. We achieve this by building Taylor's approximation for gradually changing known cardiovascular relationships between input and output (e.g., sensor measurements to blood pressure) and incorporating this approximation into our proposed neural network training. The effectiveness of the framework is demonstrated through a case study: continuous cuffless BP estimation from time series bioimpedance data. We show that by using PINNs over the state-of-the-art time series models tested on the same datasets, we retain high correlations (systolic: 0.90, diastolic: 0.89) and low error (systolic: 1.3 ± 7.6 mmHg, diastolic: 0.6 ± 6.4 mmHg) while reducing the amount of ground truth training data on average by a factor of 15. This could be helpful in developing future AI algorithms to help interpret pervasive physiologic data using minimal amount of training data.

Continuous cuffless blood pressure monitoring with a wearable ring bioimpedance device.

Smart rings provide unique opportunities for continuous physiological measurement. They are easy to wear, provide little burden in comparison to other smart wearables, are suitable for nocturnal settings, and can be sized to provide ideal contact between the sensors and the skin at all times. Continuous measuring of blood pressure (BP) provides essential diagnostic and prognostic value for cardiovascular health management. However, conventional ambulatory BP measurement devices operate using an inflating cuff that is bulky, intrusive, and impractical for frequent or continuous measurements. We introduce ring-shaped bioimpedance sensors leveraging the deep tissue sensing ability of bioimpedance while introducing no sensitivity to skin tones, unlike optical modalities. We integrate unique human finger finite element model with exhaustive experimental data from participants and derive optimum design parameters for electrode placement and sizes that yields highest sensitivity to arterial volumetric changes, with no discrimination against varying skin tones. BP is constructed using machine learning algorithms. The ring sensors are used to estimate arterial BP showing peak correlations of 0.81, and low error (systolic BP: 0.11 ± 5.27 mmHg, diastolic BP: 0.11 ± 3.87 mmHg) for >2000 data points and wide BP ranges (systolic: 89-213 mmHg and diastolic: 42-122 mmHg), highlighting the significant potential use of bioimpedance ring for accurate and continuous estimation of BP.

Physics-Informed Neural Networks for Modeling Physiological Time Series: A Case Study with Continuous Blood Pressure.

The bold vision of AI-driven pervasive physiological monitoring, through the proliferation of off-the-shelf wearables that began a decade ago, has created immense opportunities to extract actionable information for precision medicine. These AI algorithms model the input-output relationships of a system that, in many cases, exhibits complex nature and personalization requirements. A particular example is cuffless blood pressure estimation using wearable bioimpedance. However, these algorithms need to be trained with a significant amount of ground truth data. In the context of biomedical applications, collecting ground truth data, particularly at the personalized level is challenging, burdensome, and in some cases infeasible. Our objective is to establish physics-informed neural network (PINN) models for physiological time series data that would reduce reliance on ground truth information. We achieve this by building Taylor's approximation for the gradually changing known cardiovascular relationships between input and output (e.g., sensor measurements to blood pressure) and incorporating this approximation into our proposed neural network training. The effectiveness of the framework is demonstrated through a case study: continuous cuffless BP estimation from time series bioimpedance data. We show that by using PINNs over the state-of-the-art time series regression models tested on the same datasets, we retain a high correlation (systolic: 0.90, diastolic: 0.89) and low error (systolic: 1.3 ± 7.6 mmHg, diastolic: 0.6 ± 6.4 mmHg) while reducing the amount of ground truth training data on average by a factor of 15. This could be helpful in developing future AI algorithms to help interpret pervasive physiologic data using minimal amount of training data.

Graphene e-tattoos for unobstructive ambulatory electrodermal activity sensing on the palm enabled by heterogeneous serpentine ribbons.

Electrodermal activity (EDA) is a popular index of mental stress. State-of-the-art EDA sensors suffer from obstructiveness on the palm or low signal fidelity off the palm. Our previous invention of sub-micron-thin imperceptible graphene e-tattoos (GET) is ideal for unobstructive EDA sensing on the palm. However, robust electrical connection between ultrathin devices and rigid circuit boards is a long missing component for ambulatory use. To minimize the well-known strain concentration at their interfaces, we propose heterogeneous serpentine ribbons (HSPR), which refer to a GET serpentine partially overlapping with a gold serpentine without added adhesive. A fifty-fold strain reduction in HSPR vs. heterogeneous straight ribbons (HSTR) has been discovered and understood. The combination of HSPR and a soft interlayer between the GET and an EDA wristband enabled ambulatory EDA monitoring on the palm in free-living conditions. A newly developed EDA event selection policy leveraging unbiased selection of phasic events validated our GET EDA sensor against gold standards.

Arterial Pulse Localization with Varying Electrode Sizes and Spacings in Wrist-Worn Bioimpedance Sensing.

Bioimpedance has emerged as a promising modality to continuously monitor hemodynamic and respiratory physiological parameters through a non-invasive skin-contact approach. Bioimpedance sensors placed at the radial zone of the volar wrist provide sensitive operation to the blood flow of the underlying radial artery. The translation of bioimpedance systems into medical-grade settings for continuous hemodynamic monitoring, however, presents challenges when constraining the necessary sensing components to a minimal form factor while maintaining sufficient accuracy and precision of measurements. Thus, it is important to understand the effects of electrode configuration on bioimpedance signals when reducing them to a wearable form factor. Previous work regarding electrode configurations in bioimpedance does not address wearable constraints, nor do they focus on electrodes viable for wearable applications. In this study, we present empirical evidence of the effects of dry silver electrode sizes and spacings on the specificity and sensitivity of a wrist-worn bioimpedance sensor array. We found that wrist-worn bioimpedance systems for hemodynamic monitoring would benefit from reduced injection electrode spacings (up to a 392% increase in signal amplitude with a 50% decrease in spacing), increased sensing electrode spacings, and decreased electrode surface areas. Clinical Relevance - The work directly contributes towards the development of cuffless continuous blood pressure monitors with applications in clinical and ambulatory settings.

Blood Pressure Estimation using a Single Channel Bio-Impedance Ring Sensor.

The demand for non-obtrusive, accurate, and continuous blood pressure (BP) monitoring systems is becoming more prevalent with the realization of its significance in preventable cardiovascular disease (CVD) globally. Current cuff-based standards are bulky, uncomfortable, and are limited to discrete recording periods. Wearable sensor technologies such as those using optical photoplethysmography (PPG) have been used to develop blood pressure estimation models through a variety of methods. However, this technology falls short as optical based systems have bias favoring lighter skin tones and lower body fat compositions. Bioimpedance (Bio-Z) is a capable modality of sensing arterial blood flow without implicit inadvertent bias towards individuals. In this paper we propose a ring-based bioimpedance system to capture arterial blood flow from the digital artery of the finger. The ring design provides a more compact wearable device utilizing only a single Bio-Z channel, making it a familiar fit to individuals. Post-processing the acquired Bio-Z signals, we extracted 9 frequency domain features from windowed beat cycles to train subject specific regression models. Results indicate the average mean absolute errors for systolic/diastolic BP to be 4.38/3.63mmHg, consistent with AAMI standards.

Boosted-SpringDTW for Comprehensive Feature Extraction of PPG Signals.

Goal: To achieve high-quality comprehensive feature extraction from physiological signals that enables precise physiological parameter estimation despite evolving waveform morphologies. Methods: We propose Boosted-SpringDTW, a probabilistic framework that leverages dynamic time warping (DTW) and minimal domain-specific heuristics to simultaneously segment physiological signals and identify fiducial points that represent cardiac events. An automated dynamic template adapts to evolving waveform morphologies. We validate Boosted-SpringDTW performance with a benchmark PPG dataset whose morphologies include subject- and respiratory-induced variation. Results: Boosted-SpringDTW achieves precision, recall, and F1-scores over 0.96 for identifying fiducial points and mean absolute error values less than 11.41 milliseconds when estimating IBI. Conclusion: Boosted-SpringDTW improves F1-Scores compared to two baseline feature extraction algorithms by 35% on average for fiducial point identification and mean percent difference by 16% on average for IBI estimation. Significance: Precise hemodynamic parameter estimation with wearable devices enables continuous health monitoring throughout a patients' daily life.

Continuous cuffless monitoring of arterial blood pressure via graphene bioimpedance tattoos.

Continuous monitoring of arterial blood pressure (BP) in non-clinical (ambulatory) settings is essential for understanding numerous health conditions, including cardiovascular diseases. Besides their importance in medical diagnosis, ambulatory BP monitoring platforms can advance disease correlation with individual behaviour, daily habits and lifestyle, potentially enabling analysis of root causes, prognosis and disease prevention. Although conventional ambulatory BP devices exist, they are uncomfortable, bulky and intrusive. Here we introduce a wearable continuous BP monitoring platform that is based on electrical bioimpedance and leverages atomically thin, self-adhesive, lightweight and unobtrusive graphene electronic tattoos as human bioelectronic interfaces. The graphene electronic tattoos are used to monitor arterial BP for >300 min, a period tenfold longer than reported in previous studies. The BP is recorded continuously and non-invasively, with an accuracy of 0.2 ± 4.5 mm Hg for diastolic pressures and 0.2 ± 5.8 mm Hg for systolic pressures, a performance equivalent to Grade A classification.

Non-Invasive Cardiac and Respiratory Activity Assessment From Various Human Body Locations Using Bioimpedance.

OBJECTIVE: Bioimpedance sensing is a powerful technique that measures the tissue impedance and captures important physiological parameters including blood flow, lung movements, muscle contractions, body fluid shifts, and other cardiovascular parameters. This paper presents a comprehensive analysis of the modality at different arterial (ulnar, radial, tibial, and carotid arteries) and thoracic (side-rib cage and top thoracolumbar fascia) body regions and offers insights into the effectiveness of capturing various cardiac and respiratory activities. METHODS: We assess the bioimpedance performance in estimating inter-beat (IBI) and inter -breath intervals (IBrI) on six-hours of data acquired in a pilot-study from five healthy participants at rest. RESULTS: Overall, we achieve mean errors as low as 0.003 ± 0.002 and 0.67 ± 0.28 seconds for IBI and IBrI estimations, respectively. CONCLUSIONS: The results show that bioimpedance can be effectively used to monitor cardiac and respiratory activities both at limbs and upper body and demonstrate a strong potential to be adopted by wearables that aim to provide high-fidelity physiological sensing to address precision medicine needs.

A Wrist-worn Respiration Monitoring Device using Bio-Impedance.

In the US alone, 22 million individuals suffer from obstructive sleep apnea (OSA), with 80% of the cases symptoms undiagnosed. Hence, there is an unmet need to continuously and unobtrusively monitor respiration and detect possible occurrences of apnea. Recent advancements in wearable biomedical technology can enable the capture of the periodicity of the heart pressure pulse from a wrist-worn device. In this paper, we propose a bio-impedance (Bio-Z)-based respiration monitoring system. We establish close contact with the skin using gold e-tattoos with a 35 mm by 5 mm active sensing area. We extracted the respiration from the wrist Bio-Z signal leveraging three different techniques and showed that we can detect the start of each respiration beat with an average root mean square error (RMSE) less than 13% and mean error of 0.3% over five subjects.

ImpediBands: Body Coupled Bio-Impedance Patches for Physiological Sensing Proof of Concept.

Continuous and robust monitoring of physiological signals is essential in improving the diagnosis and management of cardiovascular and respiratory diseases. The state-of-the-art systems for monitoring vital signs such as heart rate, heart rate variability, respiration rate, and other hemodynamic and respiratory parameters use often bulky and obtrusive systems or depend on wearables with limited sensing methods based on repetitive properties of the signals rather than the morphology. Moreover, multiple devices and modalities are typically needed for capturing various vital signs simultaneously. In this paper, we introduce ImpediBands: small-sized distributed smart bio-impedance (Bio-Z) patches, where the communication between the patches is established through the human body, eliminating the need for electrical wires that would create a common potential point between sensors. We use ImpediBands to collect instantaneous measurements from multiple locations over the chest at the same time. We propose a blind source separation (BSS) technique based on the second-order blind identification (SOBI) followed by signal reconstruction to extract heart and lung activities from the Bio-Z signals. Using the separated source signals, we demonstrate the performance of our system via providing strong confidence in the estimation of heart and respiration rates with low RMSE (HRRMSE = 0.579 beats per minute, RRRMSE = 0.285 breaths per minute), and high correlation coefficients (rHR = 0.948, rRR = 0.921).

Measurement of Chest Physiological Signals using Wirelessly Coupled Bio-Impedance Patches.

Continuous monitoring of respiration and heart pressure pulse waves generated by lung and heart movements is essential in the diagnosis and management of cardiovascular and lung diseases. Traditional methods in measuring these physiological signals are not convenient for long-term monitoring during daily activities and sleeping due to their use of long wires and/or face masks, and leading to patients having to spend long duration in clinics while undergoing supervised monitoring. In this paper, we present a new method to measure global chest physiological signals using small-sized smart band-age like patches placed at different locations of the chest. The introduced patches communicate with each other using human body and detect small variations in the bio-Impedance (Bio-Z) depending on the mechanical heart and lung movements, where one patch injects an AC current from one side of the chest and several Bio-Z sensors measure the voltage difference across different locations of the chest. In order to prevent usage of long wires and increase convenience for wearable applications, electrical connection between current injection and Bio-Z sensing patches are eliminated. Independent component analysis is used to separate sources of physiological observations and improve accuracy of the system. We show that the presented system can successfully detect respiration rate and heart rate with an average error of 4.86% (less than 1 breath per minute) and 1.86% (less than 2 beats per minute) respectively tested on 6 healthy subjects over 6 minutes of data collection for each subject.