Research On Level Set Medical Image Segmentation Technology Based On Cluster Optimization

Image segmentation refers to the skills of dividing an image into non-overlapping objects of interest according to its inherent characteristics,such as gray level,texture and contrast.At present,image segmentation skills have been used for most kinds of images,such as terrain image,satellite image,medical image and so on.In medical treatment,image segmentation can help diagnose tumors and lesions in any part of the brain or body by MRI or PET scanning and other medical imaging techniques.It's mainly applies to brain,cardiac ventricular,abdominal organ and cell segmentation in biological images.In such applications,the results of segmentation are generally derived for quantitative measurements and biomarkers in subsequent diagnosis and treatment plans to diagnose diseases and track disease progress.This kind of research projects is one significant and challenging research direction in computer-aided diagnosis and pattern recognition.So far,there is no kinds of universal approach of medical image segmentation algorithm.The image segmentation results are generally influenced by lots of elements,such as image intensity inhomogeneity,spatial characteristics of strong correlation,high gray closeness of content and different soft tissues,which are emergent issues to be dealed with in image segmentation research.Therefore,for the issues of medical image intensity non-uniformity and limited spatial resolution,an optimized level set image segmentation approach has been presented.Firstly,the spectral clustering method on account of Nystrom strategy is applied to evaluate the random sampling record points,after that the k-means method is applied to cluster vector space.The clustering results are applied to the improved variable region fitting energy model,and the iterative evolution outputs the final experimental results.Besides,in addition to the spectral clustering approach on account of the graph theory,applying fuzzy set theory to image segmentation is also in great demand.Among them,the fuzzy C-means clustering approach classifies images by grouping similar record points in characteristic space,which leads to a outstanding experimental result.However,when the image is noisy or distorted,fuzzy C-means clustering misclassifies noisy pixels due to its irregular characteristic data.For medical images susceptible to noise pollution and weak edges,a spatial intuitionistic fuzzy clustering based on ant colony algorithm optimization is introduced to segment medical images.Firstly,an adaptive ant colony optimization algorithm is applied to obtain the foremost centroid and the amount of clusters,and then the intuitionistic fuzzy clustering associated with spatial information is applied to distinguish image pixels,and then the experiment process is accomplished.Finally,for confirming the experimental effect of the approved algorithm,a level set image segmentation platform based on MATLAB is proposed,which provides a convenient and operable simulation platform for researchers.By reading ordinary images and Dicom medical format images,the platform demonstrates the application of level set algorithm to the evolutionary iterative process of image segmentation.For the similar image,the expression of the method is estimated by visual effect,similarity coefficient amongst the final effect boundary and the reference mask,segmentation time and so on.