Figure 4 | Scientific Reports

Figure 4

From: An integrated algorithm for single lead electrocardiogram signal analysis using deep learning with 12-lead data

Figure 4

Trained model parameters setting for testing CinC-2017 dataset (a) A binary cutoff was selected for each isolated lead trained model (.h5), where the testing performance showed the highest possible accuracy. (b) The appearance of single lead ECG (SL-ECG) testing samples was evaluated on the clinical lead I trained model, with decision boundaries tested at various thresholds (i.e., 0.40, 0.44) to optimally separate the classes. The decision boundary refers to a threshold, used to convert the model’s continuous output, i.e., probability score into binary class label.

Back to article page