Fig. 5
From: Three-stage segmentation of lung region from CT images using deep neural networks
The three successive steps of the proposed lung region segmentation. Preprocessing: The CNN-1 identifies only the CT slices that contains lung region followed by quality evaluation. Processing: The identified CT slice is fed to a UNET-1 which converts the slice to a binary image. Postprocessing: Each connected components in the binary image are fed to CNN-2 and UNET-2 to extract only connected components associated with the lung region and for contour refinement, respectively, before the segmented image is displayed