Auto-Segmentation Of Liver And Kidney Based On CT Image

Cancer has become one of the most harmful disease in the past decades and will keep its harm in the future. In order to cure cancer, radiation therapy is an essential step and extracting the organs from medical images accurately is the key step in3D liver positioning and making radiation therapy plans. We still have some problems in auto segmentation of liver and kidney:the similar gray levels between liver and other organs make it difficult to identify the contours, livers show great anatomical variations among different patients; two kidneys should be segmented at the same time, water, air and other noise in kidneys also increase the difficulty of segmentation. In recent years, many researchers in the world have focused on medical image segmentation and proposed various techniques. Due to the advantage research on the Accurate/Advanced Radiation Therapy System (ARTS), the auto segmentation of liver and kidney was developed based on this platform, including three aspects:auto segmentation of liver, auto segmentation of kidney, manual segmentation tools.Segmentation methods based on region growing algorithm were widely used in applications due to the mature research and low time complexity. For the complex characteristics of abdominal medical image, original region growing algorithm is not able to segment liver and kidney accurately. In order to control region growing to fit liver and kidney while smoothing contour at the same time, an improved region growing algorithm was developed.In this article, region growing algorithm was improved based on three different aspects:seed area choose based on prior experience and liver characters; dynamic optimization of region growing rule based on edge detection results by Canny operator for liver and adaptive threshold results for kidney; contour post processing based on flood fill, morphological algorithm and curve fitting.Results of auto segmentation were not always accurate, manual segmentation tools were developed to solve this, including three useful tools:pen, roll and eraser. At last, CT image and its contour after segmentation were visualized on screen based on VTK.Several series of abdominal CT slice were used to test the algorithm and the results were compared with manual segmentation results by doctor, liver was segmented accurately on most CT slices in very short time. The experiment results showed that-the algorithm works well in controlling region growing and smoothing contour, it can improve the accuracy of liver auto segmentation and ensure speed at the same time.