AN EFFICIENT ALGORITHMIC SOLUTION FOR AUTOMATIC SEGMENTATION OF LUNGS FROM CT IMAGES

Abstract

A novel technique for lung segmentation from input Computed Tomography (CT) images using optimal thresholding was developed. Initially, the CT image was segmented by optimal thresholding. The lung volume was obtained using connected component labeling method by removing irrelevant information. The resultant image contained holes which were filled by morphological operations. A novel technique to separate the lungs was introduced which effectively separated the right and left lungs. Finally, the lung contour was smoothed by rolling ball algorithm to include any juxta pleural nodules. The proposed system was evaluated using 84 scans of publicly available dataset Lung Image Database Consortium (LIDC). The proposed system achieved an overlap measure of 0.985 and the root mean square difference between the proposed method and ground truth was 0.47 mm.