The Segmentation And Application Of Modified Clustering Method On Medical Image

Medical image segmentation is a medical image processing to the more critical step in the process of medical image analysis. Medical image segmentation can be accurate for medical workers to disease diagnosis and analysis put forward effective treatment, provide reliable theory basis for clinical medical research, allowing researchers to follow-up work for accurate decisions.In this paper, it is mainly studies about the semi-supervised learning theory and the fuzzy clustering algorithm and the semi-supervised fuzzy clustering algorithm, and it also used the classical unsupervised fuzzy c-means clustering algorithm, a semi-supervised fuzzy c-means clustering algorithm in MRI image segmentation experiment has been carried out, in view of the current algorithm in the practical difficulties and problems in MRI image segmentation, in this paper proposes a fuzzy c-means clustering algorithm, basing on a semi-supervised novel clustering algorithm, and successfully applied to the nuclear magnetic resonance image,and in this paper, we obtain the following results:(1) The fuzzy clustering algorithm and the integrated application is the method basing on semi-supervised learning theory. In classic unsupervised fuzzy c-means clustering based on the introduction of a semi-supervised learning algorithm, and through the experimental data set to join the semi-supervised algorithm after effective verification, the feasibility of the experiment proved that after joining a semi-supervised algorithms, a new algorithm in clustering accuracy and speed than the traditional fuzzy c-means clustering algorithm has significantly improved, proved a semi-supervised fuzzy c-means clustering algorithm, which is better than the classical unsupervised clustering effect on fuzzy c-means clustering algorithm is better.(2) Because of the complexity of the medical magnetic resonance imaging (MRI) image, a semi-supervised fuzzy clustering analysis was carried out on the MRI images in a large number of data samples using trace mark information study, the degradation problems here, given the degenerative, this paper proposes a novel monitoring algorithm solves the half supervision that cannot use tag information of fuzzy c-means clustering problems, but for the nuclear magnetic resonance image segmentation process, so they don’t lose a semi-supervised learning, and the new algorithm has been succeeded in medical magnetic resonance image segmentation, the experimental results from the segmentation accuracy or at the time of the algorithm is better than the first, and has a strong ability to resist the noise in noise environment, finally reached a conclusion that this novel a semi-supervised clustering algorithm in MRI image segmentation obtained better robustness and segmentation accuracy, fast convergence speed, the number of iterations, at least to verify the feasibility and effectiveness of the proposed algorithm.