Research On Weak Boundary Tracking And Lung Tumor Segmentation

Lung cancer is one of the malignant tumors that threaten human life and health.Computed tomography(CT)plays a vital role in the early screening,diagnosis and clinical treatment of lung cancer.However,the information in CT image is rich,so how to detect the region of interest and accurately segment the tumor are still difficult things in the field of computer-aided diagnosis.For CT image processing and tumor segmentation,the characteristics of different tumors are different,especially,some tumors might crowd out the surrounding tissues as they grow,resulting in part of them not within the normal lung parenchyma contour,which greatly increases the difficulty of tumor segmentation.In this paper,by analyzing and summarizing the characteristics of different tumors,corresponding methods are designed for each type of tumors such as lung parenchyma segmentation,repair of concave contours of lung parenchyma and lung tumor segmentation.The main research contents are as follows:(1)In order to improve the efficiency of images processing,an adaptive detection method is designed for the tumor region of interest.By mapping the region of interest containing the tumors to the original CT image or to the complete lung parenchyma image,global image processing is reduced to local image processing,which is beneficial to the later period of lung parenchymal contour repair and tumor segmentation.Firstly,the gray-scale histogram is combined with the Otsu method to achieve rapid initial extraction of lung parenchyma,and the rolling ball method is used to process them to achieve the initial repair operation of the contour of the lung parenchyma with depression;then the tumor localization is carried out based on the gray information of the image;finally,the window size of the region of interest is adjusted adaptively according to the tumor size information(2)In order to solve the adhesion problem of peripheral tumor,a weak boundary extraction method is designed based on directional edge tracking algorithm.The adhesion behavior of peripheral tumor can lead to the formation of weak boundary,and the way of approximating the normal lung parenchyma contour to the weak boundary of the tumor has an error that is magnified by the loss of a portion of the tumor area when the tumor invade the adherent tissue.In this paper,based on the gray information of image which is the most basic,combining the gradient information and the location of the tumor,starting from the strong boundary of the lung parenchymas,the lung parenchymal boundary is gradually searched until the weak boundary of the tumor is tracked,at the same time,using the maximum inhibition to achieve the boundary correction,finally,the weak bounday of the tumor is extracted,to achieve the purpose of repairing the lung parenchymal contour with the closest weak boundary of the real tumor.(3)To study the segmentation methods of different types of tumors,an adaptive region growth algorithm combining morphological processing is designed.The number and distribution of the seed points in the region growing algorithm affects the accuracy of the segmentation,so in this article,morphology processing is used to select seed points at the edge of the tumor area inside in order to achieve the automatic segmentation of the tumor,and the edge after inflation is used to be the limit of region growing,to reduce the growth leak situation,to raise the segmentation accuracy.