Interpretable heartbeat classification using local model-agnostic explanations on ECGs

Treatment and prevention of cardiovascular diseases often rely on Electrocardiogram (ECG) interpretation. Dependent the physician's variability, ECG interpretation is subjective prone to errors. Machine learning models are developed used support doctors; however, their lack interpretability stands as one main drawbacks widespread operation. This paper focuses an Explainable Artificial Intelligence (XAI) solution make heartbeat classification more explainable using several state-of-the-art model-agnostic methods. We introduce a high-level conceptual framework for time series propose original method that adds temporal dependency between samples series' derivative. The results were validated in MIT-BIH arrhythmia dataset: we performed performance's analysis evaluate whether explanations fit model's behaviour; employed 1-D Jaccard's index compare subsequences extracted from interpretable model XAI methods used. Our show use raw signal its derivative includes promote explanation. A small but informative user study concludes this potential visual produced by our being adopted real-world clinical settings, either diagnostic aids or training resource.