Effectual lung segmentation for CAD systems using CT scan images

The paper depicts an innovative and efficient methodology for the effective segmentation of lungs based on the true composition of the thoracic CT image. The method comprises of three phases: image enhancement of lung CT, elliptical thorax extraction and lungs segmentation. The proposed technique covers up the deficiencies present in classical segmentation techniques, where any abnormality present at the edges of the lung gets removed, being considered as part of the thorax area by the algorithm. The effectiveness of the proposed segmentation technique lies in the precise identification of the thorax and lung tissue. The method has been extensively tested, by processing datasets from six benign and three malignant cases containing 964 CT slices. The segmentation technique has also been tested against the manual traced borders by the radiologist. The technique produces comparable results with respect to experts. The approach presented is also quite efficient, performance wise and computationally as well.