Research On Image Segmentation Algorithm And Its Application In Lung Nodule Segmentation

Segmentation of pulmonary nodules in CT images of the lungs is a very important step in the computer-aided detection of lungs.Segmentation of pulmonary nodules plays a crucial role in the diagnosis and treatment of lung cancer.However,since the pulmonary nodules have boundary ambiguity in the CT image of the lung,it is difficult to obtain good result for the segmentation of the pulmonar y nodules.In addition,there are pseudo nodules in the lung nodule segmentation image,in order to better assist the doctor’s diagnosis,it is necessary to remove the pseudo nodules in the lung nodule segmentation image.Aiming at these problems,in this paper we carried out in-depth research on image segmentation algorithm and applied it to the segmentation of pulmonary nodules.The main contents are as follows:For the problem of boundary ambiguity of lung nodules in lung CT images,we study the image segmentation algorithms based on secondary segmentation theory,and analyze various classical segmentation algorithms deeply.We find that in terms of segmentation of images with blurred boundaries between target and background and edge extraction of images,OTSU and KSW algorithms are complementary.Based on the theory of quadratic segmentation,two improved algorithms are researched by combining these two algorithms: secondary segmentation algorithm based on OTSU and KSW and secondary segmentation algorithm based on misclassified regions.For the problem of pseudo nodules in the segmentation image of lung nodules,in this paper we analyze the characteristics of the pseudo-nodules in the lung nodules segmentation image,and find that these pseudo nodules are misclassified due to their gray value are similar to the gray value of nodules.And conventional filtering methods cannot effectively remove these pseudo nodules.In order to solve this problem,in this paper a pseudo target removal algorithm based on deci sion tree is researched.Based on the improved methods,in this paper,the proposed secondary segmentation algorithm based on misclassified regions is used to segment the Lung parenchyma,and then the proposed secondary segmentation algorithm based on OTSU and KSW is applied to segment the pulmonary nodules,and finally the pseudo target removal algorithm based on decision tree is used to remove the pseudo nodule regions in the lung nodule segmentation image.The experimental results show that the proposed algorithms can achieve good results.