基于Gmac模型的肺结节分割-计算机软件与理论专业论文.docxVIP

Abstract Lung cancer is the cancer with the highest mortality rate, early diagnosis can significantly improve the patients’ survivability. Pulmonary nodule is the major sign of early developing lung cancer, thus judging whether the pulmonary nodule is benign or malignant based on CT-scan image has become the focus of Computer-Aided Diagnosis(CAD) methodology, and accurate segmentation of pulmonary nodule is the first problem needs to be solved. Two major problems exist for current pulmonary nodule segmentation methods, accuracy and efficiency. Simple methods yield result fast but have low accuracy, while complicated approaches can give good results, but execution efficiency is low. Gmac(Global minimum active contour) model based pulmonary nodule segmentation method can solve the contradiction quite well, it gives promising results and give them fast. The method is based on the difference between the nodule region and the background, it’s composed of three phases, image preprocessing, Gmac-based segmentation and post processing. During the preprocessing phase, the type of the nodule is identified and different processing is performed based on the type. In segmentation phase, foreground and background mean value manipulation is applied during the iteration. During post processing, purification and de-conglutination is performed on the segmented image to acquire a more precise result. The method is tested on the Lung Imaging Database Consortium(LIDC) database and the images provided by the Second Affiliated Hospital of Soochow University. And the three Overlap criterion(overlap rate, true positive rate and false positive rate) is applied on the segmentation result of the first data set in LIDC database, also a comparison is made with other published methods that use the LIDC data base to evaluate their results. The tested result of the three criteria are 84%, 71% and 24%. Certain improvement in performance is observed, and meanwhile the time consumption is g