Research Of The Segmentation Method For Liver And Liver Tumor In CT Images Based On A Unified Level Set Framework

Medical image analysis plays an important role in computer-aided diagnosis.Medical image segmentation is one of the cores of medical image analysis,whose purpose is to identify the anatomical structures of human organs from the surrounding tissues.Computed tomography(CT)images provide pictures of anatomical structures with high definition and high signal-to-noise ratio for clinical diagnosis in a non-invasion way.In the area of CT based clinical hepatic diagnosis,accurate and reliable segmentation of liver and liver tumor is essential for the follow-up treatment planning,evaluation and computer-assisted surgery.However,in current clinical practice,manual delineation of liver and liver tumor on each slice is still typically performed by radiologists,which could obtain the arguably most accurate segmentation results,but is time-consuming,tedious,and laborious,and introduces inter-observer variability.Additionally,due to the blurry edges,low level of contrast and intensity inhomogeneity characterizing the CT images,accurate segmentation of liver and liver tumor is regarded as difficult work,and the segmentation method is a research highlight in the area of medical image processing.In this article,we present a method for liver segmentation and a method for liver tumor segmentation in CT images respectively.The two methods are mainly grounded on a novel unified level set method(LSM),which incorporates both the region information and the edge information of image to evolve the contour.This level set framework is more resistant to edge leakage than the single-information driven LSMs for liver segmentation and surpasses many other segmentation methods for liver tumor segmentation.For liver segmentation,a hybrid image preprocessing method consisting of an anisotropy filter,a scale-specific gradient filter,a nonlinear grayscale converter and a customized binarization is proposed first to convert an input CT image into a binary image.And the condition provided by the binary image helps the region-growing to overcome the sensitivity to initial seed point location setting and growth threshold setting.Then with manual setting of a few seed points on the obtained binary image,the following region-growing is performed to extract a rough liver region.The unified level set is proposed at last to refine the coarse segmentation result.The proposed method was evaluated with a total of 40 abdominal sequences from two public datasets.SLIVER07 and 3Dircadb.The validation results show that the proposed method could obtain expected segmentation results,and required less interaction than many other state-of-the-art semiautomatic methods.For liver tumor segmentation,a local intensity clustering based LSM coupled with hidden Markov random field and the expectation-maximization(HMRF-EM)algorithm is applied to construct an enhanced edge indicator for the unified LSM.With this improvement,compared with many other segmentation models,the unified LSM could reach the segmentation results closer to the gold standards,even for complex tumors.The proposed method was validated on a total of 125 tumors from two public datasets,3Dircadb and MIDAS.The validation results of MIDAS dataset show the proposed method was robust to different gold standards.And the validation results of 3Dircadb dataset indicate that the proposed method was also competitive with other state-of-the-art methods in both accuracy and efficiency.Finally,we integrate the proposed liver segmentation method and liver tumor segmentation method,and develope a user program based on the development environment of MATLAB.This user program shows the advantages of simplicity,consistency and feedback.In addition,all the 40 livers and 125 liver tumors were used to test the user program.And the test results show that the developed program could realize accurate and reliable segmentation of liver and liver tumor.