uPVC-Net: A Universal Premature Ventricular Contraction Detection Deep Learning Algorithm

2025-06-12

AI Summary

Premature ventricular contractions (PVCs) exhibit substantial morphological variability in single-lead electrocardiograms (ECGs) due to differences in lead placement, acquisition devices (e.g., Holter monitors vs. wearable patches), and population heterogeneity—severely limiting the out-of-distribution (OOD) generalizability of existing detection models. To address this, we propose a universal deep learning framework for PVC detection in single-lead ECGs. Our approach introduces a novel multi-source, multi-lead collaborative training strategy and integrates a lightweight, time-series–optimized neural architecture. This design significantly enhances robustness across device types, demographic groups, and lead configurations. Evaluated on multiple independent test sets, our method achieves AUCs of 97.8%–99.1%, including 99.1% on wearable single-lead data—substantially outperforming state-of-the-art methods. This work establishes a new paradigm for reliable arrhythmia screening in low-resource, highly heterogeneous clinical and remote monitoring settings.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

Introduction: Premature Ventricular Contractions (PVCs) are common cardiac arrhythmias originating from the ventricles. Accurate detection remains challenging due to variability in electrocardiogram (ECG) waveforms caused by differences in lead placement, recording conditions, and population demographics. Methods: We developed uPVC-Net, a universal deep learning model to detect PVCs from any single-lead ECG recordings. The model is developed on four independent ECG datasets comprising a total of 8.3 million beats collected from Holter monitors and a modern wearable ECG patch. uPVC-Net employs a custom architecture and a multi-source, multi-lead training strategy. For each experiment, one dataset is held out to evaluate out-of-distribution (OOD) generalization. Results: uPVC-Net achieved an AUC between 97.8% and 99.1% on the held-out datasets. Notably, performance on wearable single-lead ECG data reached an AUC of 99.1%. Conclusion: uPVC-Net exhibits strong generalization across diverse lead configurations and populations, highlighting its potential for robust, real-world clinical deployment.

Problem

• Detect PVCs from single-lead ECG recordings universally
• Address variability in ECG waveforms due to diverse conditions
• Improve generalization across different lead configurations and populations

Innovation

• Universal deep learning for PVC detection
• Multi-source multi-lead training strategy
• High AUC on wearable single-lead ECG