Diagnosing as Cardiologists Do: ECG Agents with Doctor-Grounded Priors for Clinical Reasoning Across Diseases and Populations

2026-08-09

AI Summary

Existing ECG-based intelligent diagnostic systems struggle to model the expert interpretation process, leading to a disconnect between waveform measurements and clinical diagnosis. This work proposes LuminaECG, a novel framework that, for the first time, translates cardiologists’ structured ECG reading protocols into learnable visual priors. By rendering ECGs on standard grid paper, explicitly annotating P/QRS/T wave boundaries, and encoding these segments with color as visual measurement primitives, the method integrates a 2-billion-parameter vision–language model with low-rank supervised fine-tuning to enable measurement-driven diagnostic reasoning without architectural modifications. Evaluated on the CODE-test benchmark, LuminaECG achieves clinically acceptable performance, significantly improving both waveform measurement precision and diagnostic accuracy, generating reports enriched with prognostic information, and demonstrating strong zero-shot generalization across diverse populations and geographic regions.

Abstract

Cardiologists interpret electrocardiograms by localizing waveform components, measuring rhythm and interval patterns, and translating these structured observations into diagnostic evidence. Whether this expert reading process can serve as an effective prior for ECG agents remains unclear. To address this question, we introduce LuminaECG, a clinically structured ECG reasoning framework that reformulates ECG interpretation as measurement-grounded visual reading. ECG signals are rendered on standard electrocardiographic grid paper to preserve the spatial and scale cues used in clinical reading. P-wave, QRS-complex, and T-wave boundaries are explicitly delineated, and color-coded segmentation decomposes the waveform into discrete visual measurement primitives. A general 2B vision-language backbone is then trained with low-rank supervised fine-tuning to associate these primitives with diagnostic reasoning, without architectural modification. Across open, proprietary, and ECG-specialist zero-shot baselines, LuminaECG improves both waveform measurement and diagnostic recovery. It reaches a clinically meaningful reader tier on the CODE-test benchmark, transfers across geographically diverse ECG datasets without retraining, and generates reports whose structure contains an emergent prognostic signal. These findings suggest that effective ECG agents require not only larger models, but supervision that preserves the alignment between measurable waveform evidence and clinical knowledge.

Problem

• ECG interpretation
• clinical reasoning
• cardiologist priors
• diagnostic evidence
• waveform measurement

Innovation

• ECG interpretation
• clinical reasoning
• vision-language model
• measurement-grounded representation
• zero-shot transfer