Research On Segmentation Algorithms Of Pulmonary Suspected Lesions From CT Images

The mortality rate for lung cancer is higher than that for other kinds of cancers around the world. At the same time, it appears that the rate has been steadily increasing. Recently mass screening based on helical computed tomography (CT) images has become popular, and it has become the focus of lung cancer screening strategies. CT will provide great numbers of data sets. Since the majority of screening cases are normal, diagnostic reading errors may be hard to avoid. Therefore, it is necessary to develop a computer-aided diagnosis (CAD) system to help radiologists with the interpretation.Based on analyzing lots of relative literatures at home and abroad, three different segmentation algorithms are proposed. Firstly, Matlab software is used for threshold segmentation, image noise removal, calculating perimeter, area and degrees of the round. The pulmonary nodule detection method based on shape characteristics is proposed. Secondly, the Gaussian-template which is applied to template matching should be constructed according to the size of ROI, and then classified nodules are extracted from the region of interests by computing correlation coefficient.A template-matching technique for detecting nodules is proposed. Thirdly, the fuzzy clustering algorithm was studied in medical images segment. The fuzzy c-means clustering was used to segment the clinical CT images in lung. The grey image clustering adopted in this thesis, based on the histogram of pixels as the swatch, separated the CT image into several different parts, while character comparability exists in analogical areas, character otherness exists in disparate areas, so that, the nodules in lung were seperated from the CT image of lung.The experimentation proves that the algorithm can make better lung nodule detection.