Engineering digital biomarkers of interstitial glucose from noninvasive smartwatches

Examining the challenges of blood pressure estimation via photoplethysmogram

The use of observed wearable sensor data (e.g., photoplethysmograms [PPG]) to infer health measures (e.g., glucose level or blood pressure) is a very active area of research. Such technology can have a significant impact on health screening, chronic disease management and remote monitoring. A common approach is to collect sensor data and corresponding labels from a clinical grade device (e.g., blood pressure cuff) and train deep learning models to map one to the other. Although well intentioned, this approach often ignores a principled analysis of whether the input sensor data have enough information to predict the desired metric. We analyze the task of predicting blood pressure from PPG pulse wave analysis. Our review of the prior work reveals that many papers fall prey to data leakage and unrealistic constraints on the task and preprocessing steps. We propose a set of tools to help determine if the input signal in question (e.g., PPG) is indeed a good predictor of the desired label (e.g., blood pressure). Using our proposed tools, we found that blood pressure prediction using PPG has a high multi-valued mapping factor of 33.2% and low mutual information of 9.8%. In comparison, heart rate prediction using PPG, a well-established task, has a very low multi-valued mapping factor of 0.75% and high mutual information of 87.7%. We argue that these results provide a more realistic representation of the current progress toward the goal of wearable blood pressure measurement via PPG pulse wave analysis. For code, see our project page: https://github.com/lirus7/PPG-BP-Analysis.

Wearable Data From Subjects Playing Super Mario, Taking University Exams, or Performing Physical Exercise Help Detect Acute Mood Disorder Episodes via Self-Supervised Learning: Prospective, Exploratory, Observational Study

BACKGROUND

Personal sensing, leveraging data passively and near-continuously collected with wearables from patients in their ecological environment, is a promising paradigm to monitor mood disorders (MDs), a major determinant of the worldwide disease burden. However, collecting and annotating wearable data is resource intensive. Studies of this kind can thus typically afford to recruit only a few dozen patients. This constitutes one of the major obstacles to applying modern supervised machine learning techniques to MD detection.

OBJECTIVE

In this paper, we overcame this data bottleneck and advanced the detection of acute MD episodes from wearables' data on the back of recent advances in self-supervised learning (SSL). This approach leverages unlabeled data to learn representations during pretraining, subsequently exploited for a supervised task.

METHODS

We collected open access data sets recording with the Empatica E4 wristband spanning different, unrelated to MD monitoring, personal sensing tasks-from emotion recognition in Super Mario players to stress detection in undergraduates-and devised a preprocessing pipeline performing on-/off-body detection, sleep/wake detection, segmentation, and (optionally) feature extraction. With 161 E4-recorded subjects, we introduced E4SelfLearning, the largest-to-date open access collection, and its preprocessing pipeline. We developed a novel E4-tailored transformer (E4mer) architecture, serving as the blueprint for both SSL and fully supervised learning; we assessed whether and under which conditions self-supervised pretraining led to an improvement over fully supervised baselines (ie, the fully supervised E4mer and pre-deep learning algorithms) in detecting acute MD episodes from recording segments taken in 64 (n=32, 50%, acute, n=32, 50%, stable) patients.

RESULTS

SSL significantly outperformed fully supervised pipelines using either our novel E4mer or extreme gradient boosting (XGBoost): n=3353 (81.23%) against n=3110 (75.35%; E4mer) and n=2973 (72.02%; XGBoost) correctly classified recording segments from a total of 4128 segments. SSL performance was strongly associated with the specific surrogate task used for pretraining, as well as with unlabeled data availability.

CONCLUSIONS

We showed that SSL, a paradigm where a model is pretrained on unlabeled data with no need for human annotations before deployment on the supervised target task of interest, helps overcome the annotation bottleneck; the choice of the pretraining surrogate task and the size of unlabeled data for pretraining are key determinants of SSL success. We introduced E4mer, which can be used for SSL, and shared the E4SelfLearning collection, along with its preprocessing pipeline, which can foster and expedite future research into SSL for personal sensing.

Multimodal In-Vehicle Hypoglycemia Warning for Drivers With Type 1 Diabetes: Design and Evaluation in Simulated and Real-World Driving

BACKGROUND

Hypoglycemia threatens cognitive function and driving safety. Previous research investigated in-vehicle voice assistants as hypoglycemia warnings. However, they could startle drivers. To address this, we combine voice warnings with ambient LEDs.

OBJECTIVE

The study assesses the effect of in-vehicle multimodal warning on emotional reaction and technology acceptance among drivers with type 1 diabetes.

METHODS

Two studies were conducted, one in simulated driving and the other in real-world driving. A quasi-experimental design included 2 independent variables (blood glucose phase and warning modality) and 1 main dependent variable (emotional reaction). Blood glucose was manipulated via intravenous catheters, and warning modality was manipulated by combining a tablet voice warning app and LEDs. Emotional reaction was measured physiologically via skin conductance response and subjectively with the Affective Slider and tested with a mixed-effect linear model. Secondary outcomes included self-reported technology acceptance. Participants were recruited from Bern University Hospital, Switzerland.

RESULTS

The simulated and real-world driving studies involved 9 and 10 participants with type 1 diabetes, respectively. Both studies showed significant results in self-reported emotional reactions (P<.001). In simulated driving, neither warning modality nor blood glucose phase significantly affected self-reported arousal, but in real-world driving, both did (F2,68=4.3; P<.05 and F2,76=4.1; P=.03). Warning modality affected self-reported valence in simulated driving (F2,68=3.9; P<.05), while blood glucose phase affected it in real-world driving (F2,76=9.3; P<.001). Skin conductance response did not yield significant results neither in the simulated driving study (modality: F2,68=2.46; P=.09, blood glucose phase: F2,68=0.3; P=.74), nor in the real-world driving study (modality: F2,76=0.8; P=.47, blood glucose phase: F2,76=0.7; P=.5). In both simulated and real-world driving studies, the voice+LED warning modality was the most effective (simulated: mean 3.38, SD 1.06 and real-world: mean 3.5, SD 0.71) and urgent (simulated: mean 3.12, SD 0.64 and real-world: mean 3.6, SD 0.52). Annoyance varied across settings. The standard warning modality was the least effective (simulated: mean 2.25, SD 1.16 and real-world: mean 3.3, SD 1.06) and urgent (simulated: mean 1.88, SD 1.55 and real-world: mean 2.6, SD 1.26) and the most annoying (simulated: mean 2.25, SD 1.16 and real-world: mean 1.7, SD 0.95). In terms of preference, the voice warning modality outperformed the standard warning modality. In simulated driving, the voice+LED warning modality (mean rank 1.5, SD rank 0.82) was preferred over the voice (mean rank 2.2, SD rank 0.6) and standard (mean rank 2.4, SD rank 0.81) warning modalities, while in real-world driving, the voice+LED and voice warning modalities were equally preferred (mean rank 1.8, SD rank 0.79) to the standard warning modality (mean rank 2.4, SD rank 0.84).

CONCLUSIONS

Despite the mixed results, this paper highlights the potential of implementing voice assistant-based health warnings in cars and advocates for multimodal alerts to enhance hypoglycemia management while driving.

TRIAL REGISTRATION

ClinicalTrials.gov NCT05183191; https://classic.clinicaltrials.gov/ct2/show/NCT05183191, ClinicalTrials.gov NCT05308095; https://classic.clinicaltrials.gov/ct2/show/NCT05308095.

The 2023 wearable photoplethysmography roadmap

Photoplethysmography is a key sensing technology which is used in wearable devices such as smartwatches and fitness trackers. Currently, photoplethysmography sensors are used to monitor physiological parameters including heart rate and heart rhythm, and to track activities like sleep and exercise. Yet, wearable photoplethysmography has potential to provide much more information on health and wellbeing, which could inform clinical decision making. This Roadmap outlines directions for research and development to realise the full potential of wearable photoplethysmography. Experts discuss key topics within the areas of sensor design, signal processing, clinical applications, and research directions. Their perspectives provide valuable guidance to researchers developing wearable photoplethysmography technology.

Uncovering personalized glucose responses and circadian rhythms from multiple wearable biosensors with Bayesian dynamical modeling

Wearable biosensors and smartphone applications can measure physiological variables over multiple days in free-living conditions. We measure food and drink ingestion, glucose dynamics, physical activity, heart rate (HR), and heart rate variability (HRV) in 25 healthy participants over 14 days. We develop a Bayesian inference framework to learn personal parameters that quantify circadian rhythms and physiological responses to external stressors. Modeling the effects of ingestion events on glucose levels reveals that slower glucose decay kinetics elicit larger postprandial glucose spikes, and we uncover a circadian baseline rhythm for glucose with high amplitudes in some individuals. Physical activity and circadian rhythms explain as much as 40%-65% of the HR variance, whereas the variance explained for HRV is more heterogeneous across individuals. A more complex model incorporating activity, HR, and HRV explains up to 15% of additional glucose variability, highlighting the relevance of integrating multiple biosensors to better predict glucose dynamics.

AI4FoodDB: a database for personalized e-Health nutrition and lifestyle through wearable devices and artificial intelligence

The increasing prevalence of diet-related diseases calls for an improvement in nutritional advice. Personalized nutrition aims to solve this problem by adapting dietary and lifestyle guidelines to the unique circumstances of each individual. With the latest advances in technology and data science, researchers can now automatically collect and analyze large amounts of data from a variety of sources, including wearable and smart devices. By combining these diverse data, more comprehensive insights of the human body and its diseases can be achieved. However, there are still major challenges to overcome, including the need for more robust data and standardization of methodologies for better subject monitoring and assessment. Here, we present the AI4Food database (AI4FoodDB), which gathers data from a nutritional weight loss intervention monitoring 100 overweight and obese participants during 1 month. Data acquisition involved manual traditional approaches, novel digital methods and the collection of biological samples, obtaining: (i) biological samples at the beginning and the end of the intervention, (ii) anthropometric measurements every 2 weeks, (iii) lifestyle and nutritional questionnaires at two different time points and (iv) continuous digital measurements for 2 weeks. To the best of our knowledge, AI4FoodDB is the first public database that centralizes food images, wearable sensors, validated questionnaires and biological samples from the same intervention. AI4FoodDB thus has immense potential for fostering the advancement of automatic and novel artificial intelligence techniques in the field of personalized care. Moreover, the collected information will yield valuable insights into the relationships between different variables and health outcomes, allowing researchers to generate and test new hypotheses, identify novel biomarkers and digital endpoints, and explore how different lifestyle, biological and digital factors impact health. The aim of this article is to describe the datasets included in AI4FoodDB and to outline the potential that they hold for precision health research. Database URL https://github.com/AI4Food/AI4FoodDB.

In-ear infrasonic hemodynography with a digital health device for cardiovascular monitoring using the human audiome

Human bodily mechanisms and functions produce low-frequency vibrations. Our ability to perceive these vibrations is limited by our range of hearing. However, in-ear infrasonic hemodynography (IH) can measure low-frequency vibrations (<20 Hz) created by vital organs as an acoustic waveform. This is captured using a technology that can be embedded into wearable devices such as in-ear headphones. IH can acquire sound signals that travel within arteries, fluids, bones, and muscles in proximity to the ear canal, allowing for measurements of an individual's unique audiome. We describe the heart rate and heart rhythm results obtained in time-series analysis of the in-ear IH data taken simultaneously with ECG recordings in two dedicated clinical studies. We demonstrate a high correlation (r = 0.99) between IH and ECG acquired interbeat interval and heart rate measurements and show that IH can continuously monitor physiological changes in heart rate induced by various breathing exercises. We also show that IH can differentiate between atrial fibrillation and sinus rhythm with performance similar to ECG. The results represent a demonstration of IH capabilities to deliver accurate heart rate and heart rhythm measurements comparable to ECG, in a wearable form factor. The development of IH shows promise for monitoring acoustic imprints of the human body that will enable new real-time applications in cardiovascular health that are continuous and noninvasive.

Medicine 2032: The future of cardiovascular disease prevention with machine learning and digital health technology

Machine learning (ML) refers to computational algorithms that iteratively improve their ability to recognize patterns in data. The digitization of our healthcare infrastructure is generating an abundance of data from electronic health records, imaging, wearables, and sensors that can be analyzed by ML algorithms to generate personalized risk assessments and promote guideline-directed medical management. ML's strength in generating insights from complex medical data to guide clinical decisions must be balanced with the potential to adversely affect patient privacy, safety, health equity, and clinical interpretability. This review provides a primer on key advances in ML for cardiovascular disease prevention and how they may impact clinical practice.

Digital Biomarkers for Personalized Nutrition: Predicting Meal Moments and Interstitial Glucose with Non-Invasive, Wearable Technologies

Digital health technologies may support the management and prevention of disease through personalized lifestyle interventions. Wearables and smartphones are increasingly used to continuously monitor health and disease in everyday life, targeting health maintenance. Here, we aim to demonstrate the potential of wearables and smartphones to (1) detect eating moments and (2) predict and explain individual glucose levels in healthy individuals, ultimately supporting health self-management. Twenty-four individuals collected continuous data from interstitial glucose monitoring, food logging, activity, and sleep tracking over 14 days. We demonstrated the use of continuous glucose monitoring and activity tracking in detecting eating moments with a prediction model showing an accuracy of 92.3% (87.2-96%) and 76.8% (74.3-81.2%) in the training and test datasets, respectively. Additionally, we showed the prediction of glucose peaks from food logging, activity tracking, and sleep monitoring with an overall mean absolute error of 0.32 (+/-0.04) mmol/L for the training data and 0.62 (+/-0.15) mmol/L for the test data. With Shapley additive explanations, the personal lifestyle elements important for predicting individual glucose peaks were identified, providing a basis for personalized lifestyle advice. Pending further validation of these digital biomarkers, they show promise in supporting the prevention and management of type 2 diabetes through personalized lifestyle recommendations.

Translational gaps and opportunities for medical wearables in digital health

A confluence of advances in biosensor technologies, enhancements in health care delivery mechanisms, and improvements in machine learning, together with an increased awareness of remote patient monitoring, has accelerated the impact of digital health across nearly every medical discipline. Medical grade wearables-noninvasive, on-body sensors operating with clinical accuracy-will play an increasingly central role in medicine by providing continuous, cost-effective measurement and interpretation of physiological data relevant to patient status and disease trajectory, both inside and outside of established health care settings. Here, we review current digital health technologies and highlight critical gaps to clinical translation and adoption.

Considerations for Analyzing and Interpreting Data from Biometric Monitoring Technologies in Clinical Trials

BACKGROUND

The proliferation and increasing maturity of biometric monitoring technologies allow clinical investigators to measure the health status of trial participants in a more holistic manner, especially outside of traditional clinical settings. This includes capturing meaningful aspects of health in daily living and a more granular and objective manner compared to traditional tools in clinical settings.

SUMMARY

Within multidisciplinary teams, statisticians and data scientists are increasingly involved in clinical trials that incorporate digital clinical measures. They are called upon to provide input into trial planning, generation of evidence on the clinical validity of novel clinical measures, and evaluation of the adequacy of existing evidence. Analysis objectives related to demonstrating clinical validity of novel clinical measures differ from typical objectives related to demonstrating safety and efficacy of therapeutic interventions using established measures which statisticians are most familiar with.

KEY MESSAGES

This paper discusses key considerations for generating evidence for clinical validity through the lens of the type and intended use of a clinical measure. This paper also briefly discusses the regulatory pathways through which clinical validity evidence may be reviewed and highlights challenges that investigators may encounter while dealing with data from biometric monitoring technologies.

A computational framework for discovering digital biomarkers of glycemic control

Digital biomarkers can radically transform the standard of care for chronic conditions that are complex to manage. In this work, we propose a scalable computational framework for discovering digital biomarkers of glycemic control. As a feasibility study, we leveraged over 79,000 days of digital data to define objective features, model the impact of each feature, classify glycemic control, and identify the most impactful digital biomarkers. Our research shows that glycemic control varies by age group, and was worse in the youngest population of subjects between the ages of 2–14. In addition, digital biomarkers like prior-day time above range and prior-day time in range, as well as total daily bolus and total daily basal were most predictive of impending glycemic control. With a combination of the top-ranked digital biomarkers, we achieved an average F1 score of 82.4% and 89.7% for classifying next-day glycemic control across two unique datasets.

End-to-end design of wearable sensors

Wearable devices provide an alternative pathway to clinical diagnostics by exploiting various physical, chemical and biological sensors to mine physiological (biophysical and/or biochemical) information in real time (preferably, continuously) and in a non-invasive or minimally invasive manner. These sensors can be worn in the form of glasses, jewellery, face masks, wristwatches, fitness bands, tattoo-like devices, bandages or other patches, and textiles. Wearables such as smartwatches have already proved their capability for the early detection and monitoring of the progression and treatment of various diseases, such as COVID-19 and Parkinson disease, through biophysical signals. Next-generation wearable sensors that enable the multimodal and/or multiplexed measurement of physical parameters and biochemical markers in real time and continuously could be a transformative technology for diagnostics, allowing for high-resolution and time-resolved historical recording of the health status of an individual. In this Review, we examine the building blocks of such wearable sensors, including the substrate materials, sensing mechanisms, power modules and decision-making units, by reflecting on the recent developments in the materials, engineering and data science of these components. Finally, we synthesize current trends in the field to provide predictions for the future trajectory of wearable sensors.

A Scalable Risk Scoring System for COVID-19 Inpatients Based on Consumer-grade Wearables: Statistical Analysis and Model Development

Background

To provide effective care for inpatients with COVID-19, clinical practitioners need systems that monitor patient health and subsequently allow for risk scoring. Existing approaches for risk scoring in patients with COVID-19 focus primarily on intensive care units (ICUs) with specialized medical measurement devices but not on hospital general wards.

Objective

In this paper, we aim to develop a risk score for inpatients with COVID-19 in general wards based on consumer-grade wearables (smartwatches).

Methods

Patients wore consumer-grade wearables to record physiological measurements, such as the heart rate (HR), heart rate variability (HRV), and respiration frequency (RF). Based on Bayesian survival analysis, we validated the association between these measurements and patient outcomes (ie, discharge or ICU admission). To build our risk score, we generated a low-dimensional representation of the physiological features. Subsequently, a pooled ordinal regression with time-dependent covariates inferred the probability of either hospital discharge or ICU admission. We evaluated the predictive performance of our developed system for risk scoring in a single-center, prospective study based on 40 inpatients with COVID-19 in a general ward of a tertiary referral center in Switzerland.

Results

First, Bayesian survival analysis showed that physiological measurements from consumer-grade wearables are significantly associated with patient outcomes (ie, discharge or ICU admission). Second, our risk score achieved a time-dependent area under the receiver operating characteristic curve (AUROC) of 0.73-0.90 based on leave-one-subject-out cross-validation.

Conclusions

Our results demonstrate the effectiveness of consumer-grade wearables for risk scoring in inpatients with COVID-19. Due to their low cost and ease of use, consumer-grade wearables could enable a scalable monitoring system.

Trial Registration

Clinicaltrials.gov NCT04357834; https://www.clinicaltrials.gov/ct2/show/NCT04357834

Non-invasively accuracy enhanced blood glucose sensor using shallow dense neural networks with NIR monitoring and medical features

Non-invasive and accurate method for continuous blood glucose monitoring, the self-testing of blood glucose is in quest for better diagnosis, control and the management of diabetes mellitus (DM). Therefore, this study reports a multiple photonic band near-infrared (mbNIR) sensor augmented with personalized medical features (PMF) in Shallow Dense Neural Networks (SDNN) for the precise, inexpensive and pain free blood glucose determination. Datasets collected from 401 blood samples were randomized and trained with ten-fold validation. Additionally, a cohort of 234 individuals not included in the model training set were investigated to evaluate the performance of the model. The model achieved the accuracy of 97.8% along with 96.0% precision, 94.8% sensitivity and 98.7% specificity for DM classification based on a diagnosis threshold of 126 mg/dL for diabetes in fasting blood glucose. For non-invasive real-time blood glucose monitoring, the model exhibited ± 15% error with 95% confidence interval and the detection limit of 60–400 mg/dL, as validated with the standard hexokinase enzymatic method for glucose estimation. In conclusion, this proposed mbNIR based SDNN model with PMF is highly accurate and computationally cheaper compared to similar previous works using complex neural network. Some groups proposed using complicated mixed types of sensors to improve noninvasive glucose prediction accuracy; however, the accuracy gain over the complexity and costs of the systems harvested is still in questioned (Geng et al. in Sci Rep 7:12650, 2017). None of previous works report on accuracy enhancement of NIR/NN using PMF. Therefore, the proposed SDNN over PMF/mbNIR is an extremely promising candidate for the non-invasive real-time blood glucose monitoring with less complexity and pain-free.

Innovative new model predicts glucose levels without poking or prodding

With the prevalence of type II diabetes rising rapidly it has become increasingly apparent that something must be done to stem the tide. While pharmaceutical treatments aimed at lowering average blood sugar are an important tool in this endeavor, it is equally (if not more) important to motivate patients to make healthy diet and exercise choices. Recent advances in non-invasive glucose monitoring suggest that real-time patient feedback may soon be available to help guide daily patient decision-making.