Application of score-weighted class activation mapping (Score-CAM) on one dimensional convolutional neural networks (1D-CNN).

Application of Score-CAM on 1D-CNN classifying the electrocardiogram (ECG). For the intepretability, the model was trained for classifying normal ECG and ECG with Premature Ventricular Complex (PVC).

Dataset

• The publicly available dataset, PTB-XL; Wagner, P., Strodthoff, N., Bousseljot, RD. et al. PTB-XL, a large publicly available electrocardiography dataset. Sci Data 7, 154 (2020).
• One must apply for the credential access to obtain and use the data on PhysioNet.

Results

Accuracy

loss: 0.029, Acc: 0.99

Example visualization

The highlighted background indicates CAM values above 90th percentile of all the CAM value. The algorithm seems to capture where the PVC occurs.