Lung segmentation with improved graph cuts on chest CT images

Abstract

Lung segmentation is often performed as a preprocessing step on chest Computed Tomography (CT) images because it is important for identifying lung diseases in clinical evaluation. Hence, researches on lung segmentation have received much attention. In this paper, we propose a new lung segmentation method based on an improved graph cuts algorithm from the energy function. First, the lung CT images is modeled with Gaussian mixture models (GMMs), and the optimized distribution parameters can be obtained with expectation maximization (EM) algorithm. With that parameters, we can construct the improved regional penalty item in the graph cuts energy function. Second, considering the image edge information, the Sobel operator is adopted to detect and extract the lung image edges, and the lung image edges information is used to improve the boundary penalty item of graph cuts energy function. Finally, the improved energy function of graph cuts algorithm is obtained, then the corresponding graph is created, and lung is segmented with the minimum cut theory. The experiments demonstrate that the proposed method is very accurate and efficient for lung segmentation.