Figure 5
The 2-dimensional model performance evaluation metrics are presented with leads on the x-axis and accuracy (%) on the y-axis. (a) The Area Under the ROC Curve (AUROC) for both validation (PTB-XL, CPSC-2018) and testing (CinC-2017) datasets is shown. (b) Accuracy and loss curves for the validation (PTB-XL, CPSC-2018) and testing (CinC-2017) sets are provided. A notable observation is that the loss and lead deflection are correlated with testing accuracy. For instance, lead aVR, which exhibits negative deflection during training, resulted in lower testing accuracy for single lead ECG in the CinC-2017 dataset.
