ECG-R1: Protocol-Guided and Modality-Agnostic MLLM for Reliable ECG Interpretation

Electrocardiography (ECG) serves as an indispensable diagnostic tool in clinical practice, yet existing multimodal large language models (MLLMs) remain unreliable for ECG interpretation, often producing plausible but clinically incorrect analyses. To address this, we propose ECG-R1, the first reasoning ECG MLLM designed for reliable ECG interpretation via three innovations. First, we construct the interpretation corpus using \textit{Protocol-Guided Instruction Data Generation}, grounding interpretation in measurable ECG features and monograph-defined quantitative thresholds and diagnostic logic. Second, we present a modality-decoupled architecture with \textit{Interleaved Modality Dropout} to improve robustness and cross-modal consistency when either the ECG signal or ECG image is missing. Third, we present \textit{Reinforcement Learning with ECG Diagnostic Evidence Rewards} to strengthen evidence-grounded ECG interpretation. Additionally, we systematically evaluate the ECG interpretation capabilities of proprietary, open-source, and medical MLLMs, and provide the first quantitative evidence that severe hallucinations are widespread, suggesting that the public should not directly trust these outputs without independent verification. Code is available at \href{https://github.com/PKUDigitalHealth/ECG-R1}{here}.

Electrocardiograms (ECGs) are widely used to assess heart health, but current AI systems that combine images and language can produce ECG explanations that sound convincing while containing medical errors. This paper introduces ECG-R1, an AI system designed to make ECG interpretation more reliable. ECG-R1 follows a step-by-step ECG reading process based on clinical guidance and uses measurable ECG information, such as heart rhythm, intervals, and waveform patterns, to support its conclusions. It can work with ECG images, raw ECG signals, or both, and is trained to give consistent answers when one type of ECG data is unavailable. The system is also trained to reward explanations that connect each diagnostic conclusion to specific ECG evidence. In experiments, ECG-R1 produced more accurate diagnoses, more useful explanations, and more stable results than existing general medical AI systems and previous ECG-focused models. These results suggest that ECG-R1 is a step toward safer AI-assisted ECG interpretation, while final clinical decisions should remain under qualified medical supervision.