Separating Expert Retention from Autonomous Source Inference in Raw-ECG-Replay-Free Continual ECG Deployment

2026-07-01

AI Summary

This work addresses the challenge of continual deployment of multi-source electrocardiogram (ECG) models under strict privacy constraints that prohibit raw data replay and lack source identifiers. To this end, the authors propose a label-free incremental learning framework that freezes a pretrained ECGFounder backbone and assigns each new data source an independent linear expert head. A lightweight MLP-based router infers the likely source using only stored features and labels, enabling source-aware adaptation without explicit source annotations. During inference, robustness is enhanced through a validation-calibrated top-2 expert fusion strategy. Evaluated on four benchmark datasets—including CPSC and PTB-XL—the method achieves a Macro-F1 score of 0.7782 in the absence of source labels, with performance gaps of merely 0.0111–0.0133 compared to an oracle model that has access to true source identities, significantly outperforming existing baselines.

Abstract

In multi-source ECG deployment, models may need to incorporate new data sources when earlier raw ECGs cannot be retained or replayed. Freezing a pretrained backbone and assigning each source an isolated classifier prevents parameter interference, but deployment still requires selecting an expert when source metadata are unavailable. We study this distinction through \ours{}, an incremental expert bank built on frozen 1024-dimensional ECGFounder features. Each arriving domain adds a balanced-softmax linear expert, while a lightweight router is fitted only on retained training features and domain labels from sources observed so far. A validation-calibrated margin rule fuses the two most likely experts instead of committing to a single routed expert. On CPSC, PTB-XL, Georgia, and Chapman-Shaoxing, source-aware expert selection reaches $0.7915\pm0.0036$ Macro-F1 and a matched offline independent-head reference reaches $0.7885\pm0.0009$, supporting strong source-aware expert retention. Without source IDs, an MLP router reaches $0.7756\pm0.0027$ and top-2 margin fusion reaches $0.7782\pm0.0022$. The top-2 gain over hard MLP routing is small ($+0.0026$), with a 95\% confidence interval from paired bootstrap that includes zero. Across three domain orders, the top-2-to-oracle gap remains $0.0111$--$0.0133$, identifying autonomous source inference as the main remaining bottleneck. No raw ECGs are replayed, but frozen training features are retained for router updates; the method is therefore not memory-free.

Problem

• continual learning
• ECG deployment
• source inference
• expert selection
• raw-ECG-replay-free

Innovation

• continual learning
• expert bank
• source-free inference
• ECG classification
• feature replay