K-MERL framework improves ECG analysis through innovative multimodal learning that works with any lead configuration while leveraging LLM-extracted knowledge.

• Overcomes traditional 12-lead dependency, making the technology accessible in resource-limited settings
• Enhances signal-text alignment by using structured knowledge from medical reports
• Improves diagnostic accuracy through better representation learning
• Bridges the gap between ECG signals and complex medical language

This research significantly advances cardiac care by enabling more accurate ECG interpretation with fewer leads, potentially expanding access to quality cardiac assessment in underserved areas.

Knowledge-enhanced Multimodal ECG Representation Learning with Arbitrary-Lead Inputs