Extending Pretrained 10-Second ECG Foundation Models to Longer Horizons

2026-05-16

AI Summary

Existing pretrained electrocardiogram (ECG) foundation models are typically constrained to fixed-length 10-second segments, rendering them ill-suited for clinical long-duration ECG recordings of variable length. This limitation introduces two key challenges: input length incompatibility and inefficient temporal information aggregation. To address these issues, this work proposes a parameter-efficient, plug-and-play framework that enables effective modeling of arbitrary-length ECG sequences while keeping the original pretrained backbone frozen. By incorporating structurally compatible long-sequence processing and semantics-guided temporal modeling, the method circumvents the information loss inherent in sliding-window or naive pooling strategies. Extensive experiments demonstrate consistent and significant performance gains over baseline approaches across multiple long-duration ECG tasks, diverse datasets, and various foundation models, achieving both high accuracy and exceptional parameter efficiency.

Abstract

Electrocardiogram (ECG) foundation models pretrained on typical diagnostic 10-second ECG segments, have demonstrated strong transferability across a range of clinical applications. However, many real-world applications produce recordings that are typically longer, and are varied in duration during inference time. These 10-second models have no built-in way to combine information across time. Extending them to longer horizons introduces two challenges: structural incompatibilities arising from input-length disparities, and semantic challenges that limit meaningful temporal aggregation. We propose a parameter-efficient framework that extends pretrained ECG foundation models to longer and variable-length ECGs without retraining the backbone. Guided by a frozen pretrained 10-second model, we introduce a lightweight plug-in module that extends the model in two complementary ways: (i) structurally compatible long-sequence processing and (ii) semantically informed temporal modeling. Experiments on multiple long-horizon ECG tasks, datasets, and foundation model backbones demonstrate that our method enables robust long-horizon extension from pretrained snapshot models, consistently outperforming sliding-window and pooling-based baselines with strong parameter efficiency.

Problem

• ECG foundation models
• long-horizon ECG
• temporal aggregation
• input-length disparity
• variable-length recordings

Innovation

• ECG foundation models
• long-horizon extension
• parameter-efficient adaptation
• temporal modeling
• variable-length sequences