Automatic Lung Segmentation Based on Texture and Deep Features of HRCT Images with Interstitial Lung Disease

Algorithm 1

Training for lung ROI selection for HRCT image.

	Input: annotated images.
(1)	Step 1: preprocessing.
(2)	1.1: utilization of Wiener filter for removing the Gaussian noise of the lung area.
(3)	Step 2: landmark detection.
(4)	2.1: hand-crafted feature selection using GLCM mining the texture features.
(5)	2.2: deep feature selection using U-Net including convolution and pooling.
(6)	2.3: feature fusion combining the deep and texture features.
(7)	Step 3: ROI selection.
(8)	3.1: define the position of ROI based on the features.
(9)	if the ROI has the same edge with the previous training then
(10)	Segment the ROI.
	Output: the learned lung ROI mask.