Interpretable deep learning for automatic diagnosis of 12-lead electrocardiogram

Figures

• 

  Figure 1. Comparison of existing models for automatic diagnosis of ECG abnormalities (A) Two-stage traditional methods using feature engineering; (B) end-to-end deep learning methods.

• 

  Figure 2. Deep neural network architecture for cardiac arrhythima diagnosis Our deep neural network accepts raw ECG inputs (12 leads, duration of 30 s, sampling rate of 500 Hz), utilizes 1D CNNs to extract deep features, and outputs the prediction results for 9 diagnostic classes.

• 

  Figure 3. F1 score comparison of machine learning models and end-to-end deep learning models

• 

  Figure 4. Interpretability of the deep learning model at both the patient level and population level using SHAP values

• 

  Figure 5. Explanation of the model's prediction results for several ECG instances from different patients The features with high contribution (i.e., SHAP values) are highlighted in orange. Only the last 10 s of top 2 influential leads are displayed due to the limited space.

• 

  Figure 6. Population-level interpretation by calculating the contribution rate of ECG leads toward diagnostic classes in the 12-lead deep model