CAMEL: An ECG Language Model for Forecasting Cardiac Events

2026-02-17

AI Summary

This work addresses the limited predictive capability of existing electrocardiogram (ECG) language models, which hinders their utility in early clinical intervention. We propose the first multimodal ECG language model endowed with long-term temporal reasoning to forecast future cardiac events. The model employs a dedicated ECG encoder to achieve cross-modal alignment between physiological signals and textual descriptions, thereby introducing prospective prediction into ECG language modeling for the first time. To advance the field, we introduce a new benchmark, ECGForecastBench, and adopt LoRA-based fine-tuning combined with curriculum learning for end-to-end training across ECG classification, metric estimation, and multi-turn diagnostic reasoning tasks. Evaluated on six tasks spanning nine datasets, our model demonstrates exceptional zero-shot performance, improving the average score on ECGBench by 7.0%, surpassing fully supervised baselines by 12.4% on ECGForecastBench, and outperforming zero-shot ELM by 21.1%.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Electrocardiograms (ECG) are electrical recordings of the heart that are critical for diagnosing cardiovascular conditions. ECG language models (ELMs) have recently emerged as a promising framework for ECG classification accompanied by report generation. However, current models cannot forecast future cardiac events despite the immense clinical value for planning earlier intervention. To address this gap, we propose CAMEL, the first ELM that is capable of inference over longer signal durations which enables its forecasting capability. Our key insight is a specialized ECG encoder which enables cross-understanding of ECG signals with text. We train CAMEL using established LLM training procedures, combining LoRA adaptation with a curriculum learning pipeline. Our curriculum includes ECG classification, metrics calculations, and multi-turn conversations to elicit reasoning. CAMEL demonstrates strong zero-shot performance across 6 tasks and 9 datasets, including ECGForecastBench, a new benchmark that we introduce for forecasting arrhythmias. CAMEL is on par with or surpasses ELMs and fully supervised baselines both in- and out-of-distribution, achieving SOTA results on ECGBench (+7.0% absolute average gain) as well as ECGForecastBench (+12.4% over fully supervised models and +21.1% over zero-shot ELMs).

Problem

• ECG forecasting
• cardiac event prediction
• ECG language models
• early intervention

Innovation

• ECG language model
• cardiac event forecasting
• cross-modal encoding
• curriculum learning
• zero-shot inference