Insulin resistance (IR), a primary precursor to type 2 diabetes, is characterized by impaired insulin action in tissues1. However, diagnostic methods remain expensive and inaccessible, which hinders early intervention2,3. Here we present the WEAR-ME study, a large, remotely conducted study of IR (n = 1,165 participants; median body mass index (BMI) = 28 kg m−2, median age = 45 years, median haemoglobin A1c (HbA1c) = 5.4%) that uses time-series data from wearable devices and routine blood biomarkers to train deep neural networks against a ground-truth measure of IR (homeostatic model assessment of IR; HOMA-IR). Using a HOMA-IR cut-off of 2.9, our multimodal model achieved robust performance (area under the receiver operating characteristic curve (AUROC) = 0.80, sensitivity = 76%, specificity = 84%) with data from wearable devices, together with demographic and routine blood biomarker data. To enhance the use of time-series data from wearables, we fine-tuned a wearable foundation model (WFM) pretrained on 40 million hours of sensor data. In an independent validation cohort (n = 72), a model integrating WFM-derived representations with demographic data surpassed a demographics-only baseline (AUROC = 0.75 versus 0.66). Moreover, adding WFM-derived representations to a model with demographics, fasting glucose and a lipid panel substantially improved performance, compared with an identical model without data from wearables (AUROC = 0.88 versus 0.76). We integrate IR prediction into a large language model to contextualize the results and facilitate personalized recommendations. This work establishes a scalable, accessible framework for the early detection of metabolic risk, which could enable timely lifestyle interventions to prevent progression to type 2 diabetes. A machine-learning model that integrates data from wearable devices (such as smartwatches) with blood biomarkers and demographic data can predict whether someone has insulin resistance, enabling timely lifestyle interventions to prevent progression to type 2 diabetes.

Cardiovascular-Kidney-Metabolic (CKM) syndrome represents a growing public health crisis, yet its subclinical heterogeneity remains underexplored. Early detection of physiological deviation is critical for preventing irreversible organ damage. Here, we characterize the prevalence and interplay of CKM impairment in a US cohort (N=819) by integrating continuous wearable data with clinical biomarkers. We assessed cardiovascular, kidney via clinical biomarkers, namely Chol/HDL, eGFR, as well as metabolic health risk through Homeostatic Model Assessment of Insulin Resistance (HOMA-IR). We show that while metabolic and cardiovascular disruptions are significantly associated (r=0.26, p<0.001), early-stage kidney impairment manifests independently. Utilizing a normalized deviance score, we identified significant health impairments in 29.0

Large language models (LLMs) have emerged as powerful tools for analyzing and interpreting complex datasets. Recent studies demonstrate their potential to generate useful, personalized responses when provided with patient-specific health information that encompasses lifestyle, biomarkers, and context. As LLM-driven health applications are increasingly adopted, rigorous and efficient one-sided evaluation methodologies are crucial to ensure response quality across multiple dimensions, including accuracy, personalization, relevance and safety. However, current evaluation practices, particularly for open-ended text responses, heavily rely on human experts. This approach introduces human factors (perspectives, potential biases, inconsistencies) and is often cost-prohibitive, labor-intensive, and hinders scalability, especially in complex domains like healthcare where response assessment necessitates domain expertise and considers multifaceted patient data, which is often nuanced and diverse. In this work, we introduce Adaptive Precise Boolean rubrics: an evaluation framework that aims to streamline human and automated evaluation of open-ended questions by identifying critical gaps in model responses using a minimal set of targeted rubric questions. Our approach is based on recent work in more general evaluation settings that contrasts a smaller set of complex evaluation targets with a larger set of more precise, granular targets answerable with simple Boolean responses. We validate this approach in metabolic health, a domain encompassing diabetes, cardiovascular disease, and obesity. Our results demonstrate that Adaptive Precise Boolean rubrics yield substantially higher inter-rater agreement among both expert and non-expert human evaluators, as well as in automated assessments, compared to traditional Likert scales, while requiring approximately half the evaluation time of Likert-based methods. This enhanced efficiency and scalability, particularly through automated evaluation and non-expert contributions, paves the way for more extensive and cost-effective evaluation of LLMs in health.

Health is a fundamental pillar of human wellness, and the rapid advancements in large language models (LLMs) have driven the development of a new generation of health agents. However, the application of health agents to fulfill the diverse needs of individuals in daily non-clinical settings is underexplored. In this work, we aim to build a comprehensive personal health agent that is able to reason about multimodal data from everyday consumer wellness devices and common personal health records, and provide personalized health recommendations. To understand end-users' needs when interacting with such an assistant, we conducted an in-depth analysis of web search and health forum queries, alongside qualitative insights from users and health experts gathered through a user-centered design process. Based on these findings, we identified three major categories of consumer health needs, each of which is supported by a specialist sub-agent: (1) a data science agent that analyzes personal time-series wearable and health record data, (2) a health domain expert agent that integrates users' health and contextual data to generate accurate, personalized insights, and (3) a health coach agent that synthesizes data insights, guiding users using a specified psychological strategy and tracking users' progress. Furthermore, we propose and develop the Personal Health Agent (PHA), a multi-agent framework that enables dynamic, personalized interactions to address individual health needs. To evaluate each sub-agent and the multi-agent system, we conducted automated and human evaluations across 10 benchmark tasks, involving more than 7,000 annotations and 1,100 hours of effort from health experts and end-users. Our work represents the most comprehensive evaluation of a health agent to date and establishes a strong foundation towards the futuristic vision of a personal health agent accessible to everyone.

BACKGROUND:Wrist-worn wearables can detect irregular heart rhythms using photoplethysmography, but ECGs are required to confirm atrial fibrillation (AF). We sought to determine the frequency of a recurrent irregular heart rhythm detection (IHRD; ≥30 minutes of an irregular rhythm), estimate the potential diagnostic yield of different electrocardiographic monitoring strategies for confirming AF, and identify predictors of recurrent IHRDs. METHODS:The Fitbit Heart Study enrolled wrist-worn photoplethysmography device users without diagnosed AF. Of 455 699 participants, 1057 who wore and returned a 1-week ECG patch monitor after receiving an IHRD were analyzed. Baseline clinical data, device-derived metrics, IHRDs during follow-up, and electrocardiographic patch data were used for analysis. RESULTS:A total of 570 (53.9%) participants were aged 40 to 64 years, 422 (39.9%) were aged ≥65 years, and 510 (48.2%) were women. Median follow-up after ECG patch initiation was 80 days (interquartile range, 45-122 days). The frequency of another IHRD was 57.2% (95% CI, 53.1%-60.9%) at 3 months. After an initial IHRD, the estimated diagnostic yield for AF with a 10-second ECG was 7.6% (95% CI, 6.2%-9.0%), twice-daily 30-second ECGs over 1 week 19.0% (95% CI, 16.7%-21.2%), 24-hour monitor 17.4% (95% CI, 15.5%-19.3%), 1-week monitor 32.2% (95% CI, 29.4%-35.0%), 2-week monitor 46.8% (95% CI, 42.7%-50.8%), and 4-week monitor 60.8% (95% CI, 56.5%-65.1%). The risk of a recurrent IHRD was greater with older age (P<0.001), male sex (P=0.001), vascular disease (P=0.03), longer initial runs of consecutive IHRDs at detection (P=0.02), and less nightly sleep (P=0.03). CONCLUSIONS:Irregular heart rhythms are common after initial detection using a wrist-worn wearable device. Longer electrocardiographic monitoring periods increase the likelihood of confirming AF. REGISTRATION:URL: https://www.clinicaltrials.gov; Unique identifier: NCT04380415.

BackgroundThe early detection of respiratory infections could improve responses against outbreaks. Wearable devices can provide insights into health and well-being using longitudinal physiological signals. ObjectiveThe purpose of this study was to prospectively evaluate the performance of a consumer wearable physiology-based respiratory infection detection algorithm in health care workers. MethodsIn this study, we evaluated the performance of a previously developed system to predict the presence of COVID-19 or other upper respiratory infections. The system generates real-time alerts using physiological signals recorded from a smartwatch. Resting heart rate, respiratory rate, and heart rate variability measured during the sleeping period were used for prediction. After baseline recordings, when participants received a notification from the system, they were required to undergo testing at a Northwell Health System site. Participants were asked to self-report any positive tests during the study. The accuracy of model prediction was evaluated using respiratory infection results (laboratory results or self-reports), and postnotification surveys were used to evaluate potential confounding factors. ResultsA total of 577 participants from Northwell Health in New York were enrolled in the study between January 6, 2022, and July 20, 2022. Of these, 470 successfully completed the study, 89 did not provide sufficient physiological data to receive any prediction from the model, and 18 dropped out. Out of the 470 participants who completed the study and wore the smartwatch as required for the 16-week study duration, the algorithm generated 665 positive alerts, of which 153 (23.0%) were not acted upon to undergo testing for respiratory viruses. Across the 512 instances of positive alerts that involved a respiratory viral panel test, 63 had confirmed respiratory infection results (ie, COVID-19 or other respiratory infections detected using a polymerase chain reaction or home test) and the remaining 449 had negative upper respiratory infection test results. Across all cases, the estimated false-positive rate based on predictions per day was 2%, and the positive-predictive value ranged from 4% to 10% in this specific population, with an observed incidence rate of 198 cases per week per 100,000. Detailed examination of questionnaires filled out after receiving a positive alert revealed that physical or emotional stress events, such as intense exercise, poor sleep, stress, and excessive alcohol consumption, could cause a false-positive result. ConclusionsThe real-time alerting system provides advance warning on respiratory viral infections as well as other physical or emotional stress events that could lead to physiological signal changes. This study showed the potential of wearables with embedded alerting systems to provide information on wellness measures.

Background: Morbidity from undiagnosed atrial fibrillation (AF) may be preventable with early detection. Many consumer wearables contain optical photoplethysmography (PPG) sensors to measure pulse rate. PPG-based software algorithms that detect irregular heart rhythms may identify undiagnosed AF in large populations using wearables, but minimizing false-positive detections is essential. Methods: We performed a prospective remote clinical trial to examine a novel PPG-based algorithm for detecting undiagnosed AF from a range of wrist-worn devices. Adults aged ≥22 years in the United States without AF, using compatible wearable Fitbit devices and Android or iOS smartphones, were included. PPG data were analyzed using a novel algorithm that examines overlapping 5-minute pulse windows (tachograms). Eligible participants with an irregular heart rhythm detection (IHRD), defined as 11 consecutive irregular tachograms, were invited to schedule a telehealth visit and were mailed a 1-week ambulatory ECG patch monitor. The primary outcome was the positive predictive value of the first IHRD during ECG patch monitoring for concurrent AF. Results: A total of 455 699 participants enrolled (median age 47 years, 71% female, 73% White) between May 6 and October 1, 2020. IHRDs occurred for 4728 (1%) participants, and 2070 (4%) participants aged ≥65 years during a median of 122 (interquartile range, 110–134) days at risk for an IHRD. Among 1057 participants with an IHRD notification and subsequent analyzable ECG patch monitor, AF was present in 340 (32.2%). Of the 225 participants with another IHRD during ECG patch monitoring, 221 had concurrent AF on the ECG and 4 did not, resulting in an IHRD positive predictive value of 98.2% (95% CI, 95.5%–99.5%). For participants aged ≥65 years, the IHRD positive predictive value was 97.0% (95% CI, 91.4%–99.4%). Conclusions: A novel PPG software algorithm for wearable Fitbit devices exhibited a high positive predictive value for concurrent AF and identified participants likely to have AF on subsequent ECG patch monitoring. Wearable devices may facilitate identifying individuals with undiagnosed AF. Registration: URL: https://www.clinicaltrials.gov ; Unique identifier: NCT04380415.

Background Early detection of atrial fibrillation or Flutter (AF) may enable prevention of downstream morbidity. Consumer wrist-worn wearable technology is capable of detecting AF by identifying irregular pulse waveforms using photoplethysmography (PPG). The validity of PPG-based software algorithms for AF detection requires prospective assessment. Methods The Fitbit Heart Study (NCT04380415) is a single-arm remote clinical trial examining the validity of a novel PPG-based software algorithm for detecting AF. The proprietary Fitbit algorithm examines pulse waveform intervals during analyzable periods in which participants are sufficiently stationary. Fitbit consumers with compatible wrist-worn trackers or smartwatches were invited to participate. Enrollment began May 6, 2020 and as of October 1, 2020, 455,699 participants enrolled. Participants in whom an irregular heart rhythm was detected were invited to attend a telehealth visit and eligible participants were then mailed a one-week single lead electrocardiographic (ECG) patch monitor. The primary study objective is to assess the positive predictive value of an irregular heart rhythm detection for AF during the ECG patch monitor period. Additional objectives will examine the validity of irregular pulse tachograms during subsequent heart rhythm detections, self-reported AF diagnoses and treatments, and relations between irregular heart rhythm detections and AF episode duration and time spent in AF. Conclusions The Fitbit Heart Study is a large-scale remote clinical trial comprising a unique software algorithm for detection of AF. The study results will provide critical insights into the use of consumer wearable technology for AF detection, and for characterizing the nature of AF episodes detected using consumer-based PPG technology. (Am Heart J 2021;238:16- 26.)

OBJECTIVES:To determine whether X-ray fused with MRI (XFM) is beneficial for select transcatheter congenital heart disease interventions. BACKGROUND:Complex transcatheter interventions often require three-dimensional (3D) soft tissue imaging guidance. Fusion imaging with live X-ray fluoroscopy can potentially improve and simplify procedures. METHODS:Patients referred for select congenital heart disease interventions were prospectively enrolled. Cardiac MRI data was overlaid on live fluoroscopy for procedural guidance. Likert scale operator assessments of value were recorded. Fluoroscopy time, radiation exposure, contrast dose, and procedure time were compared to matched cases from our institutional experience. RESULTS:Forty-six patients were enrolled. Pre-catheterization, same day cardiac MRI findings indicated intervention should be deferred in nine patients. XFM-guided cardiac catheterization was performed in 37 (median age 8.7 years [0.5-63 years]; median weight 28 kg [5.6-110 kg]) with the following prespecified indications: pulmonary artery (PA) stenosis (n = 13), aortic coarctation (n = 12), conduit stenosis/insufficiency (n = 9), and ventricular septal defect (n = 3). Diagnostic catheterization showed intervention was not indicated in 12 additional cases. XFM-guided intervention was performed in the remaining 25. Fluoroscopy time was shorter for XFM-guided intervention cases compared to matched controls. There was no significant difference in radiation dose area product, contrast volume, or procedure time. Operator Likert scores indicated XFM provided useful soft tissue guidance in all cases and was never misleading. CONCLUSIONS:XFM provides operators with meaningful three-dimensional soft tissue data and reduces fluoroscopy time in select congenital heart disease interventions.

Introduction: Atrial fibrillation screening tools such portable ECG, blood pressure cuff, or smartphone photoplethysmography are single time-point measurements and are limited in detecting low to m...

Rapid application of external defibrillation, a crucial first-line therapy for ventricular fibrillation and cardiac arrest, is currently unavailable in the setting of magnetic resonance imaging (MRI), raising concerns about patient safety during MRI tests and MRI-guided procedures, particularly in patients with cardiovascular diseases. The objective of this study was to examine the feasibility and safety of defibrillation/pacing for the entire range of clinically useful shock energies inside the MRI bore and during scans, using defibrillation/pacing outside the magnet as a control. Experiments were conducted using a commercial defibrillator (LIFEPAK 20, Physio-Control, Redmond, Washington, USA) with a custom high-voltage, twisted-pair cable with two mounted resonant floating radiofrequency traps to reduce emission from the defibrillator and the MRI scanner. A total of 18 high-energy (200-360 J) defibrillation experiments were conducted in six swine on a 1.5 T MRI scanner outside the magnet bore, inside the bore, and during scanning, using adult and pediatric defibrillation pads. Defibrillation was followed by cardiac pacing (with capture) in a subset of two animals. Monitored signals included: high-fidelity temperature (0.01 °C, 10 samples/sec) under the pads and 12-lead electrocardiogram (ECG) using an MRI-compatible ECG system. Defibrillation/pacing was successful in all experiments. Temperature was higher during defibrillation inside the bore and during scanning compared with outside the bore, but the differences were small (ΔT: 0.5 and 0.7 °C, p = 0.01 and 0.04, respectively). During scans, temperature after defibrillation tended to be higher for pediatric vs. adult pads (p = 0.08). MR-image quality (signal-to-noise ratio) decreased by ~ 10% when the defibrillator was turned on. Our study demonstrates the feasibility and safety of in-bore defibrillation for the full range of defibrillation energies used in clinical practice, as well as of transcutaneous cardiac pacing inside the MRI bore. Methods for Improving MR-image quality in the presence of a working defibrillator require further study.

The hallmark of heart failure is increased blood volume. Quantitative blood volume measures are not conveniently available and are not tested in heart failure management. We assess ferumoxytol, a marketed parenteral iron supplement having a long intravascular half-life, to measure the blood volume with cardiovascular magnetic resonance (CMR).

Hemodynamic recording during interventional cardiovascular procedures is essential for procedural guidance, monitoring patient status, and collection of diagnostic information. Recent advances have made interventions guided by magnetic resonance imaging (MRI) possible and attractive in certain clinical scenarios. However, in the MRI environment, electromagnetic interference (EMI) can cause severe distortions and artifacts in acquired hemodynamic waveforms. The primary aim of this paper was to develop and validate a system to minimize EMI on electrocardiogram (ECG) and invasive blood pressure (IBP) signals. A system was developed which incorporated commercial MRI compatible ECG leads and pressure transducers, custom electronics, user interface, and adaptive signal processing. Measurements were made on pediatric patients (N = 6) during MRI-guided catheterization. Real-time interactive scanning, which is known to produce significant EMI due to fast gradient switching and varying imaging plane orientations, was selected for testing. The effectiveness of the adaptive algorithms was determined by measuring the reduction of noise peaks, amplitude of noise peaks, and false QRS triggers. During real-time gradient-intensive imaging sequences, peak noise amplitude was reduced by 80% and false QRS triggers were reduced to a median of 0. There was no detectable interference on the IBP channels. A hemodynamic recording system front-end was successfully developed and deployed, which enabled high-fidelity recording of ECG and IBP during MRI scanning. The schematics and assembly instructions are publicly available to facilitate implementation at other institutions. Researchers and clinicians are provided a critical tool in investigating and implementing MRI guided interventional cardiovascular procedures.

In X-ray fluoroscopy, static overlays are used to visualize soft tissue. We propose a system for cardiac and respiratory motion compensation of these overlays. It consists of a 3-D motion model created from real-time magnetic resonance (MR) imaging. Multiple sagittal slices are acquired and retrospectively stacked to consistent 3-D volumes. Slice stacking considers cardiac information derived from the ECG and respiratory information extracted from the images. Additionally, temporal smoothness of the stacking is enhanced. Motion is estimated from the MR volumes using deformable 3-D/3-D registration. The motion model itself is a linear direct correspondence model using the same surrogate signals as slice stacking. In X-ray fluoroscopy, only the surrogate signals need to be extracted to apply the motion model and animate the overlay in real time. For evaluation, points are manually annotated in oblique MR slices and in contrast-enhanced X-ray images. The 2-D Euclidean distance of these points is reduced from 3.85 to 2.75 mm in MR and from 3.0 to 1.8 mm in X-ray compared with the static baseline. Furthermore, the motion-compensated overlays are shown qualitatively as images and videos.

Introduction: Establishing the safety of external defibrillation inside the MRI magnet is crucial for the emerging field of MRI-guided EP procedures. Objective: To examine the feasibility and safet...

Background: Atrial Fibrillation is the most common clinically significant cardiac arrhythmia, estimated to affect between 2% ( 65 years of age) of the U.S. population. Screening and early detection of AF can help prevent associated stroke and cardiovascular disease. Commonly used monitoring devices are limited to short periods (< 4 weeks) and are usually employed in symptomatic subjects. Wrist-worn wearables equipped with photoplethysmography (PPG) could potentially aid in AF screening, as they can be easily worn by most people for extended periods of time. Objective: This study aimed to investigate the ability of wrist-based PPG to detect the presence of AF in subjects while at rest. Methods: This study was conducted with the approval of local institutional review boards, and all patients provided written, informed consent. The first dataset consisted of 9 patients with persistent AF and 13 subjects with normal sinus rhythm. For these 22 subjects, PPG and ECG data were collected while sitting stationary for 15 minutes (awake dataset). A second dataset consisted of 10 patients with persistent AF and 27 subjects with no known diagnosis of AF. These subjects were asked to wear a Fitbit wrist-band which recorded PPG and accelerometry data during sleep (sleep dataset, 73 total nights). There was no overlap between subjects in the awake and sleep datasets. Results: Data were analyzed in overlapping 1 or 5-minute windows. Pulse Rate Variability (PRV) features, PPG morphology features and accelerometer features were extracted for each window. An algorithm was trained on the awake dataset and was validated on the sleep dataset. The performance was 95.7% (98%) sensitivity (Se) and 0.8% (0.8%) False Positive Rate (FPR) for 1 min (5 min) windows. We also investigated the shortest duration of an AF episode that could be detected by using synthesized data. With 1 minute analysis windows, the algorithm was sensitive down to 40-second episodes. Using 5-minute intervals, episodes down to 3 minutes could be reliably detected. Conclusions: Wrist-bands equipped with PPG sensors can be used to detect atrial fibrillation during sleep or during awake, stationary periods. While reliable detection of short AF episodes can be challenging due to the noisy nature of PPG signals, we have demonstrated that select PPG features can enable accurate detection of AF down to 40 second periods. These findings suggest that wrist-worn devices equipped with a PPG sensor could be used to screen for AF in high-risk subjects or monitor patients post treatment, but are limited to cases where users are not moving or exercising. Additional, larger clinical studies are planned for the future to validate these promising early results and to validate the algorithm’s specificity against other forms of arrhythmia. [iproc 2017;3(1):e17]

Quantification of cardiac output and pulmonary vascular resistance (PVR) are critical components of invasive hemodynamic assessment, and can be measured concurrently with pressures using phase contrast CMR flow during real-time CMR guided cardiac catheterization.

Children with heart disease may require repeated X-Ray cardiac catheterization procedures, are more radiosensitive, and more likely to survive to experience oncologic risks of medical radiation. Cardiovascular magnetic resonance (CMR) is radiation-free and offers information about structure, function, and perfusion but not hemodynamics. We intend to perform complete radiation-free diagnostic right heart catheterization entirely using CMR fluoroscopy guidance in an unselected cohort of pediatric patients; we report the feasibility and safety.

Background Radiofrequency ablation for ventricular arrhythmias is limited by inability to visualize tissue destruction, by reversible conduction block resulting from edema surrounding lesions, and by insufficient lesion depth. We hypothesized that transcatheter needle injection of caustic agents doped with gadolinium contrast under real-time magnetic resonance imaging (MRI) could achieve deep, targeted, and irreversible myocardial ablation, which would be immediately visible.Methods and Results Under real-time MRI guidance, ethanol or acetic acid was injected into the myocardium of 8 swine using MRI-conspicuous needle catheters. Chemoablation lesions had identical geometry by in vivo and ex vivo MRI and histopathology, both immediately and after 12 (7-17) days. Ethanol caused stellate lesions with patchy areas of normal myocardium, whereas acetic acid caused homogeneous circumscribed lesions of irreversible necrosis. Ischemic cardiomyopathy was created in 10 additional swine by subselective transcoronary ethanol administration into noncontiguous territories. After 12 (8-15) days, real-time MRI-guided chemoablationwith 2 to 5 injections to create a linear lesionsuccessfully eliminated the isthmus and local abnormal voltage activities.Conclusions Real-time MRI-guided chemoablation with acetic acid enabled the intended arrhythmic substrate, whether deep or superficial, to be visualized immediately and ablated irreversibly. In an animal model of ischemic cardiomyopathy, obliteration of a conductive isthmus both anatomically and functionally and abolition of local abnormal voltage activities in areas of heterogeneous scar were feasible. This represents the first report of MRI-guided myocardial chemoablation, an approach that could improve the efficacy of arrhythmic substrate ablation in the thick ventricular myocardium.