ExECG: An Explainable AI Framework for ECG models

Accuracy isn't enough for clinical use: The paper emphasizes that while deep learning ECG models perform well, accuracy alone cannot justify clinical deployment—doctors and patients need to understand why the model made a specific diagnosis, which is essential for building trust and catching errors.

Explainability methods lack standardization: Current ECG explainability research uses inconsistent practices and reporting conventions across different studies, making it difficult for researchers to reuse findings or reproduce results, ultimately slowing progress in the field.

ExECG provides a unified framework: The framework addresses reproducibility challenges by standardizing how different explainability methods are applied to ECG models through three integrated components (data handling, explanation generation, and visualization), enabling consistent comparison across different approaches.

Interoperability matters for real-world deployment: By supporting multiple ECG formats and diverse explanation methods within a single system, ExECG demonstrates that practical AI tools for healthcare need to work seamlessly across different data types and technical approaches to be genuinely useful in clinical settings.

Transparency enables better error analysis: When AI systems explain their reasoning through standardized explainability methods, clinicians can more effectively identify when and why models fail, supporting both safety improvements and informed decision-making in patient care.