Segmentation of Regions of Interest in Lung CT Images Based on 2-D OTSU Optimized by Genetic Algorithm

摘要

It is the key and difficult step to segment and extract suspected nodular lesions from CT images for lung cancer CAD system. A segmentation method is proposed based on 2-D OTSU optimized by genetic algorithm for Regions of Interest (ROI) in thoracic CT images in this paper. The chromosome is encoded in binaryzation by gray of original image, and the populations is produced randomly, then through operations of choice, cross and mutation, the optimum threshold value is obtained. 2-D OTSU combines information of spatial neighborhood pixels with the whole image gray value, so it can remove most of noise. Genetic algorithm is used to solve the optimum threshold, it can decrease the run time greatly. Experiment results indicated that the algorithm can extract ROIs in lung CT images effectively in consideration of the balance of efficiency and quality. The segmentation method based on 2-D OTSU optimized by genetic algorithm proposed in this paper is effective for extraction of ROI in CT images.