PaperarXiv:2611.18665
From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection
Biomedical Signal Processing and Control (Jan 2027 issue), article 111520; arXiv 2611.18665v1
TL;DR
A 12-lead ECG teacher distils what it knows into a single-lead (Lead I) student for atrial fibrillation detection, with domain-adversarial training in both stages and GAN-made style variations of the single-lead signal so the student copes with new devices and datasets. Trained on six public 12-lead datasets and tested on the AliveCor single-lead PhysioNet 2017 set, it reaches macro-F1 0.646 and AF F1 0.782, ahead of the nine baselines compared. Code is public under MIT; no trained weights.
Abstract
Atrial fibrillation (AF) detection from long-term ambulatory electrocardiography (ECG) is critical because AF is a leading cause of stroke yet is frequently missed by conventional diagnostics, particularly in wearable settings that demand strong cross-domain generalization. Deep learning has improved AF identification by exploiting rhythm patterns, but methods such as BMIRC and SJTU-ECGNet still struggle to generalize across heterogeneous single-lead wearable datasets. To address this issue, we propose X-LeadGen, a two-stage cross-lead distillation framework for single-lead ECG AF detection that explicitly targets domain generalization. First, a 12-lead teacher network distills rich multi-lead information into a single-lead student through cross-lead knowledge transfer and a simulated self-distillation branch within the teacher, improving cross-lead representation alignment and AF discrimination. Second, we introduce a domain-generalization strategy that applies domain-adversarial training in both teacher and student stages, complemented by GANs-based augmentation in the student phase to learn domain-invariant features and enhance robustness to distribution shifts. Across seven public ECG datasets under real and simulated Lead-I settings, X-LeadGen consistently outperforms existing approaches in accuracy and F1-score, demonstrating strong generalization for real-world wearable AF monitoring. Our implementation is publicly available at https://github.com/Aveiro-Lin/X-LeadGen, enabling further research and validation in the ECG community.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
Datasets used
- PhysioNet/CinC Challenge 2017: AF classification from a short single-lead ECG (AliveCor)EvaluationECG
Chapman–Shaoxing–Ningbo 12-lead ECG arrhythmia database (PhysioNet)Training45,152 subjects · ECG- CPSC2018 (China Physiological Signal Challenge 2018): 12-lead ECG arrhythmia set, redistributed in PhysioNet Challenge 2020TrainingECG
PTB-XL21,799 clinical 12-lead ECGs (PhysioNet)Training18,869 subjects · ECG
How it works
Cite this paper
Lin, X., Li, Y., Shen, Y., Zhang, W., Liu, Y., Zhang, P., & Chen, B. (2027). From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection. Biomedical Signal Processing and Control, 111520. https://doi.org/10.1016/j.bspc.2026.111520
Lin, Xiaojian, Yang Li, Yaomin Shen, Wenxin Zhang, Yu Liu, Ping Zhang, and Bo Chen. 2027. "From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection." Biomedical Signal Processing and Control, 111520. https://doi.org/10.1016/j.bspc.2026.111520.
Lin, X., Li, Y., Shen, Y., Zhang, W., Liu, Y., Zhang, P. and Chen, B. (2027) 'From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection', Biomedical Signal Processing and Control, p. 111520. doi: 10.1016/j.bspc.2026.111520.
Lin, Xiaojian, et al. "From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection." Biomedical Signal Processing and Control, 2027, p. 111520. https://doi.org/10.1016/j.bspc.2026.111520.
X. Lin et al., "From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection," Biomedical Signal Processing and Control, p. 111520, 2027, doi: 10.1016/j.bspc.2026.111520.
@article{lin2027from,
title = {{From Twelve Leads to Single: Cross-Lead Knowledge Distillation for Domain-Generalized Atrial Fibrillation Detection}},
author = {Lin, Xiaojian and Li, Yang and Shen, Yaomin and Zhang, Wenxin and Liu, Yu and Zhang, Ping and Chen, Bo},
year = {2027},
journal = {Biomedical Signal Processing and Control},
pages = {111520},
doi = {10.1016/j.bspc.2026.111520},
eprint = {2611.18665},
archivePrefix = {arXiv}
}Added to the ChatPPG research hub