Interpretable multimodal zero shot ECG diagnosis via structured clinical knowledge alignment

Figures

• 

  Fig. 1. Qualitative analysis of ZETA’s interpretable predictions. Qualitative analysis of ZETA's interpretable predictions for First Degree Atrioventricular Block (1AVB), Axis Left Shift (ALS), and Wolf-Parkinson-White syndrome (WPW). Scores indicate the alignment between the ECG (sample leads shown) and expert-reviewed P/N clinical observations, providing a transparent rationale for the final diagnosis likelihood.

• 

  Fig. 2. Overview of the ZETA framework. (1) ECG Observation Generation & Curation: Large language models (LLMs) generate candidate positive and negative observations (e.g., for 1AVB), which are validated and refined by clinical experts. (2) Observation Comparison: A pre-trained multimodal model compares ECG embeddings against expert-reviewed observation embeddings, yielding aggregated similarity scores for a final likelihood score. This process provides interpretable insights by linking predictions back to structured, clinically relevant diagnostic features.

• 

  Fig. 3. Workflow for preprocessing LLM-generated clinical observations. Workflow for preprocessing LLM-generated clinical observations. Candidate observations from multiple LLMs are selected and organized into dataset-specific and cross-dataset sets for subsequent expert review.

• 

  Fig. 4. ZETA’s inference architecture. ZETA's inference architecture: encoding ECG and P/N observations, computing similarities, and reasoning via observation scores to predict condition likelihood. Background color intensity represents ECG-observation similarity, with darker colors indicating higher similarity.