Research On Partition Algorithms Of Brain Tumor MR Image

Medical image segmentation is significant in clinical diagnosis,clinicopathologic analysis and treatment. The accuracy of the segmentation isessential for doctors to judge pathogenetic condition and to make a correct diagnosisplanning. There are important practical values about the research on accuratesegmentation of brain tumors in medical image segmentation. Accurate segmentationof brain tissue is a prerequisite in many medical applied research such asradiotherapy planning, surgical planning,3d visualization and quantitativemeasurements of brain tumor structure,and so on. Due to the complexity of medicalimages, so far, there are no methods can produce a satisfactory segmentation effectfor all the medical images. The paper carries on a preliminary investigation intosegmentation technology of brain tumor MR image after getting some professionalknowledge about medical image segmentation algorithms. The mainly work has thefollowing aspects:1studies the brain tumor segmentation algorithms based on FCM algorithm.The paper uses k-means clustering and fuzzy c-means clustering algorithm tosegment the brain tumor, and analyzes the segmentation results of the twoalgorithms. K-means algorithm is very simple, but it can not produce a satisfactorysegmentation result for the edge blurring and low contrast medical image. BecauseFCM algorithm uses fuzzy set theory, it can present the tumor outline clearly whichk-means algorithm can not extract.2studies the brain tumor segmentation algorithms based on mathematicalmorphology. In order to solve the problem of over-segmentation existing in traditionalwatershed algorithm, the paper uses minima imposition technique andmarker-controlled to improve the traditional watershed algorithm. For the edgeblurring and low contrast medical image, first use top-hat and bottom-hattransformation to enhance the image contrast, then use the improved watershedsegmentation algorithm to segment the image. Compared with traditional watershed algorithm, over-segmentation phenomenon is well suppressed, achieves good effect.3studies the brain tumor segmentation algorithms based on GVF Snake model.The paper improves GVF Snake model. The method obtains the edge map based onCanny operator and the edges produced automatically through watershed algorithmwill be the initial contours of GVF Snake model. Then uses the improved GVF Snakealgorithm to segment brain tumor image, it not only avoids the complexity andsubjectivity of the initial contour obtained by artificial but also improves the iterationof the algorithm efficiency and accuracy. Finally, the paper compares thesegmentation results of the algorithms in this paper. In summary, the GVF Snakemodel proposed in this paper has the best segmentation effect.