Fig. 4.

Performance of the trained CNN for lung parenchyma segmentation. a The training accuracy. b The training loss. c The six convolutional kernels. d The receiver operating characteristic (ROC) curve. e The confusion matrix