Domain-Adaptive Arrhythmia Classification Using a Hybrid Transformer on Wearable Heart Signals

2026-05-05

AI Summary

This work addresses the performance degradation of electrocardiogram (ECG) classification models on wearable devices caused by domain shift. To mitigate this issue, the authors propose a dual-path hybrid Transformer architecture that jointly models raw ECG signals and seven heart rate variability (HRV) features, effectively integrating morphological and rhythmic statistical information. The model incorporates a parameter-free maximum mean discrepancy (MMD) loss at the representation level to align source and target domains, enabling efficient cross-domain generalization. Trained on multi-source ECG data, the proposed method achieves 95.0% F1-macro and 96.15% balanced accuracy on unseen wearable device data, exhibiting only a ~2% drop compared to in-domain performance. These results significantly outperform existing approaches, demonstrating the model’s strong robustness and cross-domain adaptability.

Abstract

Cardiovascular disease remains the leading cause of death globally, underscoring the need for effective, accessible monitoring solutions, particularly through wearable devices that enable continuous, real-time tracking of heart rhythms in home settings. However, deploying deep learning models trained on clinical electrocardiogram (ECG) datasets to wearable devices remains challenging, as differences in recording equipment, signal quality, and patient populations introduce domain shifts that degrade model performance. We propose a hybrid transformer model that processes continuous ECG signals alongside seven heart rate variability (HRV) features, where the raw signal path captures beat-level morphological patterns and the HRV path encodes rhythm regularity statistics, allowing the model to jointly leverage complementary information from both representations. To enhance the model's ability to generalize across domains, we employ representation learning techniques, including Maximum Mean Discrepancy (MMD), a non-parametric kernel-based metric that quantifies the distance between feature distributions of different domains, to align feature distributions between source and target domains, addressing the challenge of domain shifts between public datasets and wearable device data. By leveraging five public ECG datasets for training, the model learns robust, generalized representations that mitigate domain-specific biases. When tested on wearable device data with an unseen domain, the model achieved an F1-macro 95% and balanced accuracy of 96.15%. These results demonstrate minimal performance degradation, with only a 2% drop in F1-macro compared to seen-domain evaluation, highlighting the model's generalization capabilities and its potential for reliable, real-time heart monitoring applications in home and ambulatory settings.

Problem

• domain shift
• arrhythmia classification
• wearable ECG
• domain adaptation
• heart rhythm monitoring

Innovation

• Hybrid Transformer
• Domain Adaptation
• Heart Rate Variability (HRV)
• Maximum Mean Discrepancy (MMD)
• Wearable ECG