Search for: All records

Abstract Psychological stress is a key driver of short-term blood pressure (BP) elevations and cardiovascular risk, yet its moment-to-moment impact in daily life remains difficult to predict. In this longitudinal observational study, we collected multimodal data from 20 adults with self-reported hypertension, including continuous wearable-derived heart rate and activity, ecological momentary assessment (EMA) stress ratings, and ambulatory BP measurements in free-living conditions. The dataset comprised 3694 EMA responses and 3812 BP measurements collected over approximately four weeks per participant (mean 24.1 ± 8.5 days). We evaluated whether participant-specific (“personalized”) models outperform a single pooled population model. Two prediction tasks were examined: (i) prediction of near-term BP elevations from wearable signals and stress EMA responses and (ii) prediction of self-reported stress from wearable signals and BP. Across both tasks, personalized models consistently improved predictive performance. For BP prediction, personalized models achieved a mean AUROC of 0.803, exceeding the population model by 0.235, while for stress prediction they achieved a mean AUROC of 0.849, exceeding the population model by 0.208. These findings suggest that personalized wearable-based models can capture individual patterns of stress and BP dynamics, with direct implications for precision mental health assessment and just-in-time adaptive intervention design in future work. 

The development of digital instruments for mental health monitoring using biosensor data from wearable devices can enable remote, longitudinal, and objective quantitative benchmarks. To survey developments and trends in this field, we conducted a systematic review of artificial intelligence (AI) models using data from wearable biosensors to predict mental health conditions and symptoms. Following PRISMA guidelines, we identified 48 studies using a variety of wearable and smartphone biosensors including heart rate, heart rate variability (HRV), electrodermal activity/galvanic skin response (EDA/GSR), and digital proxies for biosignals such as accelerometry, location, audio, and usage metadata. We observed several technical and methodological challenges across studies in this field, including lack of ecological validity, data heterogeneity, small sample sizes, and battery drainage issues. We outline several corresponding opportunities for advancement in the field of AI-driven biosensing for mental health. 

Abstract ObjectiveMobile and ubiquitous devices enable health data collection “in a free-living environment” to support applications such as remote patient monitoring and adaptive digital interventions using machine learning (ML). Despite their potential, significant data collection challenges persist, including issues related to user compliance with reporting data, passive data consistency, and authorization. This scoping review identifies and analyzes these challenges, focusing on barriers to effective data collection. Materials and MethodsWe searched IEEE, ACM, and Web of Science for papers involving training ML models using both active and passive mobile sensing. We used the following search terms: “mobile OR ubiquitous”, “EMA”, “health”, “passive”, and “deep learning OR machine learning”. We only included papers that collected both passive and active data and excluded papers that used a pre-existing dataset. ResultsA total of 77 studies met the inclusion criteria. These studies utilized smartphones, smartwatches, wearable devices, and environmental sensors for data collection. Several studies reported challenges with participant compliance in active data collection, while passive data collection faced data consistency and authorization issues. Efforts to address these challenges were documented in some but not all studies. Using this information, we outline current challenges and corresponding opportunities for data collection in mobile sensing studies. DiscussionML techniques can reduce participant burden in active data collection by optimizing prompt timing, auto-filling responses, and minimizing prompt frequency. Simplified interfaces such as user-friendly smartwatch prompts can further improve compliance. For passive data collection, techniques such as optimization of recording times to preserve battery life and motivational techniques to encourage proper device use can increase data consistency. ConclusionMobile sensing offers opportunities for developing intelligent mobile health applications but faces data collection challenges with respect to factors such as compliance, consistency, and authorization. Innovations in ML and user interface design show promise for addressing these barriers. 

Substance use disorders affect 17.3% of Americans. Digital health solutions that use machine learning to detect substance use from wearable biosignal data can eventually pave the way for real-time digital interventions. However, difficulties in addressing severe between-subject data heterogeneity have hampered the adaptation of machine learning approaches for substance use detection, necessitating more robust technological solutions. We tested the utility of personalized machine learning using participant-specific convolutional neural networks (CNNs) enhanced with self-supervised learning (SSL) to detect drug use. In a pilot feasibility study, we collected data from 9 participants using Fitbit Charge 5 devices, supplemented by ecological momentary assessments to collect real-time labels of substance use. We implemented a baseline 1D-CNN model with traditional supervised learning and an experimental SSL-enhanced model to improve individualized feature extraction under limited label conditions. Results: Among the 9 participants, we achieved an average area under the receiver operating characteristic curve score across participants of 0.695 for the supervised CNNs and 0.729 for the SSL models. Strategic selection of an optimal threshold enabled us to optimize either sensitivity or specificity while maintaining reasonable performance for the other metric. Conclusion: These findings suggest that Fitbit data have the potential to enhance substance use monitoring systems. However, the small sample size in this study limits its generalizability to diverse populations, so we call for future research that explores SSL-powered personalization at a larger scale. 

Background: Machine learning models often use passively recorded sensor data streams as inputs to train machine learning models that predict outcomes captured through ecological momentary assessments (EMA). Despite the growth of mobile data collection, challenges in obtaining proper authorization to send notifications, receive background events, and perform background tasks persist. Objective: We investigated challenges faced by mobile sensing apps in real-world settings in order to develop design guidelines. For active data, we compared 2 prompting strategies: setup prompting, where the app requests authorization during its initial run, and contextual prompting, where authorization is requested when an event or notification occurs. Additionally, we evaluated 2 passive data collection paradigms: collection during scheduled background tasks and persistent reminders that trigger passive data collection. We investigated the following research questions (RQs): (RQ1) how do setup prompting and contextual prompting affect scheduled notification delivery and the response rate of notification-initiated EMA? (RQ2) Which authorization paradigm, setup or contextual prompting, is more successful in leading users to grant authorization to receive background events? and (RQ3) Which polling-based method, persistent reminders or scheduled background tasks, completes more background sessions?. Methods: We developed mobile sensing apps for iOS and Android devices and tested them through a 30-day user study asking college students (n=145) about their stress levels. Participants responded to a daily EMA question to test active data collection. The sensing apps collected background location events, polled for passive data with persistent reminders, and scheduled background tasks to test passive data collection.Results: For RQ1, setup and contextual prompting yielded no significant difference (ANOVA F1,144=0.0227; P=.88) in EMA compliance, with an average of 23.4 (SD 7.36) out of 30 assessments completed. However, qualitative analysis revealed that contextual prompting on iOS devices resulted in inconsistent notification deliveries. For RQ2, contextual prompting for background events was 55.5% (χ21=4.4; P=.04) more effective in gaining authorization. For RQ3, users demonstrated resistance to installing the persistent reminder, but when installed, the persistent reminder performed 226.5% more background sessions than traditional background tasks. Conclusions: We developed design guidelines for improving mobile sensing on consumer mobile devices based on our qualitative and quantitative results. Our qualitative results demonstrated that contextual prompts on iOS devices resulted in inconsistent notification deliveries, unlike setup prompting on Android devices. We therefore recommend using setup prompting for EMA when possible. We found that contextual prompting is more efficient for authorizing background events. We therefore recommend using contextual prompting for passive sensing. Finally, we conclude that developing a persistent reminder and requiring participants to install it provides an additional way to poll for sensor and user data and could improve data collection to support adaptive interventions powered by machine learning. 

While cities around the world are increasingly promoting streets and public spaces that prioritize pedestrians over vehicles, significant data gaps have made pedestrian mapping, analysis, and modeling challenging to carry out. Most cities, even in industrialized economies, still lack information about the location and connectivity of their sidewalks, making it difficult to implement research on pedestrian infrastructure and holding the technology industry back from developing accurate, location-based Apps for pedestrians, wheelchair users, street vendors, and other sidewalk users. To address this gap, we have designed and implemented an end-to-end open-source tool— Tile2Net —for extracting sidewalk, crosswalk, and footpath polygons from orthorectified aerial imagery using semantic segmentation. The segmentation model, trained on aerial imagery from Cambridge, MA, Washington DC, and New York City, offers the first open-source scene classification model for pedestrian infrastructure from sub-meter resolution aerial tiles, which can be used to generate planimetric sidewalk data in North American cities. Tile2Net also generates pedestrian networks from the resulting polygons, which can be used to prepare datasets for pedestrian routing applications. The work offers a low-cost and scalable data collection methodology for systematically generating sidewalk network datasets, where orthorectified aerial imagery is available, contributing to over-due efforts to equalize data opportunities for pedestrians, particularly in cities that lack the resources necessary to collect such data using more conventional methods. 

Recent advances regarding the interplay between ab initio calculations and metrology are reviewed, with particular emphasis on gas-based techniques used for temperature and pressure measurements. Since roughly 2010, several thermophysical quantities – in particular, virial and transport coefficients – can be computed from first principles without uncontrolled approximations and with rigorously propagated uncertainties. In the case of helium, computational results have accuracies that exceed the best experimental data by at least one order of magnitude and are suitable to be used in primary metrology. The availability of ab initio virial and transport coefficients contributed to the recent SI definition of temperature by facilitating measurements of the Boltzmann constant with unprecedented accuracy. Presently, they enable the development of primary standards of thermodynamic temperature in the range 2.5–552 K and pressure up to 7 MPa using acoustic gas thermometry, dielectric constant gas thermometry, and refractive index gas thermometry. These approaches will be reviewed, highlighting the effect of first-principles data on their accuracy. The recent advances in electronic structure calculations that enabled highly accurate solutions for the many-body interaction potentials and polarizabilities of atoms – particularly helium – will be described, together with the subsequent computational methods, most often based on quantum statistical mechanics and its path-integral formulation, that provide thermophysical properties and their uncertainties. Similar approaches for molecular systems, and their applications, are briefly discussed. Current limitations and expected future lines of research are assessed.