Abstract:Background: Although deep learning methods have shown great promise for identification of structural and functional cardiac abnormalities using electrocardiographic data, these methods are data hungry, posing a challenge for critically important tasks where ground truth labels are relatively scarce. Impaired coronary microvascular and vasomotor function is difficult to identify with standard clinical methods of cardiovascular testing such as coronary angiography and noninvasive single photon emission tomography (SPECT) myocardial perfusion imaging (MPI). Gold standard data from positron emission tomography (PET) are gaining emphasis in clinical guidelines but are expensive and only available in relatively limited centers. We hypothesized that signals embedded within resting and stress electrocardiograms (ECGs) identify individuals with microvascular and vasomotor dysfunction. Methods: We developed and pretrained a self-supervised foundation vision transformer model using a large database of unlabeled ECG waveforms (N=800,035). We then fine-tuned the foundation model for two clinical tasks: the difficult problem of identifying patients with impaired myocardial flow reserve (AI-MFR), and the relatively easier problem of detecting impaired LVEF (AI-LVEF). A second ECG database was labeled with task-specific annotations derived from quantitative PET MPI (N=4167). Diagnostic accuracy of AI predictions was tested in a holdout set of patients undergoing PET MPI (N=1031). Prognostic evaluation was performed in the PET holdout cohort, as well as independent cohorts of patients undergoing pharmacologic or exercise stress SPECT MPI (N=6635). Results: The diagnostic accuracy of AI-MFR with SSL pretraining increased significantly compared to de novo supervised training (AUROC, sensitivity, specificity: 0.758, 70.1%, 69.4% vs. 0.632, 66.1%, 57.3%, p<0.0001). SSL pretraining also produced a smaller increase in AI-LVEF accuracy (AUROC, sensitivity, specificity: 0.946, 89.4%, 85.9% vs. 0.918, 87.6%, 82.5%, p<0.02). Abnormal AI-MFR was found to be significantly associated with mortality risk in all three test cohorts (Hazard Ratio (HR) 2.61 [95% CI 1.83, 3.71], p<0.0001, PET cohort; HR 2.30 [2.03, 2.61], p<0.0001, pharmacologic stress SPECT cohort; HR 3.76 [2.36, 5.99], p<0.0001, exercise stress SPECT cohort). Conclusion: SSL pretraining of a vision transformer foundation model enabled identification of signals predictive of impaired MFR, a hallmark of microvascular and vasomotor dysfunction, and impaired LV function in resting and stress ECG waveforms. These signals are powerful predictors of prognosis in patients undergoing routine noninvasive stress testing and could enable more efficient diagnosis and management of these common conditions.

Large-scale Training of Foundation Models for Wearable Biosignals

Wearable Accelerometer Foundation Models for Health via Knowledge Distillation

Scaling Wearable Foundation Models

Toward Foundation Model for Multivariate Wearable Sensing of Physiological Signals

An Electrocardiogram Foundation Model Built on over 10 Million Recordings with External Evaluation across Multiple Domains

Learning Generalizable Physiological Representations from Large-scale Wearable Data

Foundation Models for ECG: Leveraging Hybrid Self-Supervised Learning for Advanced Cardiac Diagnostics

Abstract 19135: Deep Learning Model for Detection of Conduction Disorders and Arrhythmias From PDF Outputs of Commercially Available Wearable Devices

Self-supervised transfer learning of physiological representations from free-living wearable data

Practical intelligent diagnostic algorithm for wearable 12-lead ECG via self-supervised learning on large-scale dataset

DeepBeat: A multi-task deep learning approach to assess signal quality and arrhythmia detection in wearable devices

Multi-task deep learning for cardiac rhythm detection in wearable devices

Learning Individualized Cardiovascular Responses from Large-scale Wearable Sensors Data

Self-supervised learning for human activity recognition using 700,000 person-days of wearable data

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

ECG-FM: An Open Electrocardiogram Foundation Model

From Glucose Patterns to Health Outcomes: A Generalizable Foundation Model for Continuous Glucose Monitor Data Analysis

Biometric Contrastive Learning for Data-Efficient Deep Learning from Electrocardiographic Images

DeepHeart: Semi-Supervised Sequence Learning for Cardiovascular Risk Prediction

Learning From Alarms: A Robust Learning Approach for Accurate Photoplethysmography-Based Atrial Fibrillation Detection using Eight Million Samples Labeled with Imprecise Arrhythmia Alarms

Self-supervised deep representation learning of a foundation transformer model enabling efficient ECG-based assessment of cardiac and coronary function with limited labels