Lung segmentation method based on concavity degree of border points

Lungs segmentation is a vital step for quantitative lung CT image analysis and computer aided diagnosis. However, accurate and automated lung segmentation is necessary for recognizing lesion but it may be difficult by the presence of the abnormalities. Many lung diseases change tissue densities or resulting in intensity changes in the CT image data. So, intensity-only segmentation algorithms will not work for most pathological lung cases such as juxtapluaralnodule which attached to the chest wall and may be excluded from lungs. This paper presents an algorithm for pathological lung segmentation. Firstly, the quality of CT images is enhanced by noise removing. After that, lungs region borders have been extracted from the background and chest box fromcomputer tomography (CT) scan slices using morphological operations. In the second step, it closes the holes on the lung bordersbased on the concavity degree histogram computation ofborder points and its local maxima’s.For closing the holesleaded from abnormality, we find local maxima s point inconcavity histogram of the lung borders and then draw straightline between each pair of points which has less Euclideandistance than a predefine threshold otherwise we keep mainborder of the lung. At the end, the performanceof our proposedmethod is reported by experimental results using clinical CTimages. We validate our method by comparing our automaticsegmentation result with manually traced segmentation result.Sensitivity, specificity were calculated to evaluate the method