ExECG: A New Python Framework for Explainable AI in ECG Diagnostics

The newly developed Python framework, ExECG (Explainable AI for ECG models), seeks to establish a standardized approach to explainability in ECG diagnostics driven by deep learning. Although these models excel in tasks such as arrhythmia classification and abnormality detection, their lack of transparency poses challenges for clinical use. Current explainable AI (XAI) techniques for ECG differ in their methodologies and reporting standards, which hampers their reuse and reproducibility. ExECG introduces a three-phase pipeline: the Wrapper standardizes access to various ECG formats; the Explainer consolidates different XAI techniques under a common execution framework; and the Visualizer enables consistent comparisons across methods via a single interface. The framework is illustrated with brief examples, highlighting the necessity for justification, error analysis, and trust in clinical environments.