Research On MR Image Segmentation Algorithm Based On Cluster Analysis

MR images contain abundant information about human tissues.Using these information to effectively segment MR image can help doctors and patients understand the accurate information about the lesion area and the size,position,and shape of each organ.However,due to the limitations of time and space,the MR image always have different levels of noise,and the image has problems such as uneven gray levels and weakened boundaries,which brings great difficulties to MR image segmentation.In recent years,many scholars have devoted themselves to the research of clustering technology and expect to achieve efficient segmentation of MR image.However,most algorithms still have the disadvantages of low segmentation accuracy,high complexity,poor robustness,and lack of versatility.At present,the effective combination of multiple methods can be used to segment the MR image segmentation More granularly,so that it becomes a research hotspot.Therefore,this article draws upon a variety of methods to effectively combine,and proposes a soft subspace clustering algorithm optimized with fireworks algorithms and an improved density peak fast search clustering algorithm segmented MR images.The main work of this article is as follows:(1)The existing soft subspace clustering algorithm is susceptible to random noise when MR images are segmented,and it is easy to fall into local optimum due to poor robustness to the initial clustering center To solve these problems,this paper proposes a soft subspace algorithm for MR image clustering based on fireworks algorithm(FWASSC).Firstly,a new objective function with boundary constraints and noise clustering is designed to overcome the shortcomings of the existing algorithms that are sensitive to noise data.Next,a new method of calculating affiliation degree is proposed to find the subspace where the cluster is located quickly and accurately.Then,adaptive fireworks algorithm is introduced to the clustering process to effectively balance the local and global search,overcoming the disadvantage of falling into local optimum in the existing algorithms.We added noise to the composite image to demonstrate the anti-noise performance,compared the segmentation result of natural image with the golden section standard to verify the accuracy of the proposed algorithm,analyzed the theoretical time complexity and the actual time-consuming to prove the time performance of the algorithm.It is demonstrated that the proposed algorithm has good performance.(2)Due to the density clustering method is not sensitivity to noise,and the excellent performance of finding the non-spherical clusters,this section combines the superpixel algorithm SLIC and density clustering to realize the one-time division of superpixel objects and complete the image segmentation.After running the superpixel algorithm SLIC which is used to divide the MR image into a small number of superpixel objects,according to the detailed information to determine each superpixel objects' neighbors,and use uses the K-nearest neighbor optimized density peak fast search algorithm to clustering these superpixel objects.So we propose an improved density peak value fast search clustering algorithm to segment MR images.In order to verify anti-noise performance of the algorithm,we add the salt-noise noise to the natural image,and compare the segmentation result with the golden section standard to prove the segmentation accuracy,analyzed the theoretical time complexity and the actual time-consuming to demonstrate the time performance of the algorithm.Experiments show that this algorithm has good segmentation quality and superior performance for clinical MR images,and it is versatile.