A Fast Automatic Method of Lung Segmentation in CT Images Using Mathematical Morphology

In the Computer-Aided Detection (CAD) of lung nodules in x-ray computed tomography (CT) scans of the thorax, lung segmentation is the preliminary step. This paper proposes a fast, fully automated lung segmentation method for application in X-ray computed tomography. The segmentation method bases on mathematical morphology, which is known for its speed. The method operates slice and slice by three main steps: (1) the lung region is extracted from the CT images by gray-level thresholding, (2) the trachea or bronchi is removed from lung regions and the lung borders are smoothed by morphological operations, and (3) the left and right lungs are separated by geodesic dilation. The performance of the automated segmentation method was evaluated using 1099 computed tomography images (126 thick slice and 973 thin slice scans). The proposed method successfully segmented 95.6% of the 126 thick slice images and 98.3% of the 973 thin slice images. The fast, fully automated lung segmentation method has advantages over other methods in speed, robustness and accuracy.