What question did this study set out to answer?

To develop and evaluate a novel zero-shot framework for ECG diagnosis that enhances interpretability and aligns with clinical workflows.

January 10, 2026npj Cardiovascular Health0 citationsOpen Access

Interpretable multimodal zero shot ECG diagnosis via structured clinical knowledge alignment

JTJialu TangEindhoven University of Technology HPHuy PhamSingapore Management University ILIgnace De LathauwerRadboud University Nijmegen

Key Result

ZETA's zero-shot multimodal ECG framework enhances diagnostic interpretability by aligning ECG signals with structured clinical knowledge, showing competitive classification performance.

Key Points

To develop and evaluate a novel zero-shot framework for ECG diagnosis that enhances interpretability and aligns with clinical workflows.
Developed the ZETA framework that compares ECG signals with structured clinical observations.
Utilized a pre-trained multimodal model for alignment of ECG and text embeddings.
Employed expert validation and LLM assistance for curating clinical observations.
ZETA shows competitive performance in zero-shot classification of ECG.
Enhanced interpretability with predictions grounded in clinically relevant features was observed.
The method improves transparency and generalizability of AI in diagnosing cardiovascular diseases.

Structured PICO

Does the ZETA zero-shot multimodal framework provide accurate and interpretable ECG diagnosis compared to existing baselines?

Population

Public benchmark 12-lead ECG datasets (PTB-XL, CPSC2018, CSN, CODE-15) covering 88 diagnostic conditions.

Intervention

ZETA, a zero-shot multimodal framework for interpretable ECG diagnosis using structured clinical knowledge alignment.

Comparator

Existing zero-shot baselines (MERL21, D-BETA10), supervised DNNs, and human clinicians.

Outcome

Zero-shot classification performance (ROC AUC) and diagnostic accuracy in a human-centered study.

ZETA demonstrates competitive zero-shot ECG classification performance while providing structured, clinically interpretable rationales that can assist physician diagnosis.

Main Result

Absolute Event Rate: 0% vs 0%

Abstract

Abstract Electrocardiogram (ECG) interpretation is essential for cardiovascular disease diagnosis, but current automated systems often struggle with transparency and generalization to unseen conditions. To address this, we introduce ZETA, a zero-shot multimodal framework designed for interpretable ECG diagnosis aligned with clinical workflows. ZETA uniquely compares ECG signals against structured positive and negative clinical observations, which are curated through an LLM-assisted, expert-validated process, thereby mimicking differential diagnosis. Our approach leverages a pre-trained multimodal model to align ECG and text embeddings without disease-specific fine-tuning. Empirical evaluations demonstrate ZETA’s competitive zero-shot classification performance and, importantly, provide qualitative and quantitative evidence of enhanced interpretability, grounding predictions in specific, clinically relevant positive and negative diagnostic features. ZETA underscores the potential of aligning ECG analysis with structured clinical knowledge for building more transparent, generalizable, and trustworthy AI diagnostic systems.

Ask AI

Helpful

Bookmark

View Full Paper