Automatic lung segmentation in CT images using watershed transform

Abstract

The preprocessing step of most computer-aided diagnosis (CAD) systems for identifying the lung diseases is lung segmentation. We present a novel lung segmentation technique based on watershed transform, which is fast and accurate. Lung region is precisely marked with internal and external markers. The markers are combined with the gradient image of the original data and watershed transform is applied on the combined data to find the lung borders. Rolling ball filter is used to smooth the contour and fill the cavities while preserving the original borders. The proposed method eliminates the tasks of finding an optimal threshold and separating the attached left and right lungs, which are two common practices in most lung segmentation methods and require a significant amount of time. We have applied our new approach on several pulmonary CT images and the results reveal the speed, robustness and accuracy of this method.