A novel approach of lung segmentation on chest CT images using graph cuts

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, research on lung segmentation has received much attention. Most of the conventional methods need the post-processing just like rolling ball method or morphology method to deal with the juxtapleural lung nodules or other lesions. To find a most robust method of lung segmentation, we propose a new algorithm based on an improved graph cuts algorithm with Gaussian mixture models (GMMs) in this paper. The core method models the foreground object and background of the CT images as a GMMs, and the weight or probability that each pixel belongs to the foreground object is calculated with expectation maximization (EM) algorithm. The corresponding graph is then created using these weights on the nodes and edges. And the segmentation is completed with the minimum cut theory. Experimental results show that the proposed method is very accurate and efficient, and can directly provide explicit lung regions without any postprocessing operations even in complex scenarios.