Lung area extraction from 3D chest X-ray CT images using a shape model generated by a variable Be'zier surface

The lung area extraction from 3D chest X-ray CT images is an important function in performing computer diagnosis of lung cancer, etc. using CT images. In the conventional lung area extraction methods, when the lesions contact with the chest wall and mediastinum, there is a problem that defects will occur in the extracted lung area. In this paper, to deal with this problem, we will describe a lung area extraction method using a shape model. Concretely, the contour shape is modeled by using a Bezier surface and is made to fit for the contour surface of the individual input images by deformation manipulation based on the affine transformation and active contour model. Using this model, the aforementioned defects of the lung area due to lesions contacting with the wall should be corrected satisfactorily. In the experiments, we show, by applying the proposed technique to 3D chest X-ray CT images, that the lung area can be extracted satisfactorily in cases in which the aforementioned lesions exist. © 2003 Wiley Periodicals, Inc. Syst Comp Jpn, 34(4): 60–71, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.1201