Medical Image Clustering Method Based On Spectrum Theory Of Graph

The rapid development of medical imaging technology in recent years has led to the sustained progress of medical diagnosis technology.A large number of medical images are used for clinical diagnosis,but the current use of medical images relies on manual observation and judgment by medical personnel,which brings up some problems: on the one hand,the different perspectives of medical staff may lead to different diagnostic results,and the impact of medical personnel's own practice level may lead to the deviation of the patient's diagnosis.On the other hand,the simple artificial processing of medical images can't solve the demand for the growth of mass medical data processing.Therefore how to use medical images effectively has attracted a great deal of attention from researchers.Data mining technology can find valuable information from a large amount of data,and data mining technology has also emerged to deal with medical images.Clustering analysis plays a key role in data mining.Therefore,the main purpose of this paper is to study the characterization of human brain CT images,and make more effective use of the existing CT image resource.Especially for brain CT image clustering related research to help doctors diagnose and extract the hidden value.However Low-level semantic features(such as color,texture,etc.)are widely used to compare the similarity of images for the existing image feature representation.But doctors usually judge the similarity of medical images based on the differences between high-level semantic features(such as lesions,lesion types,etc.)in the local region.In this paper,we propose a novel multimodal features medical image clustering method which combines the medical image texture feature and morphological features.This method can solve the problem that traditional methods ignore the high-level semantic features of medical images,which leads to the poor clustering effect of medical images.Firstly,this paper proposes a texture feature fusion method to represent the global underlying semantic features of medical images,and the texture feature matrix of the image set is obtained by using the texture feature extraction method;on the other hand,the morphological description of the region of interest(ROI)is proposed as the local high-level semantics of medical images feature,and the morphological feature matrix of the image set is obtained by using the ROIextraction method and the ROI vectorization method.Secondly the similarity measure is used to calculate the similarity between the two types of features for CT images,and get similar values between images.Finally according to the spectrum theory of the graph,the multi-kernel spectral clustering method is used to learn the feature fusion weight.And the validity of the method is verified by using the Multi-Kernel spectral clustering experiment.The experiments in this paper are compared with several traditional and newest research results,and their clustering results are obtained under different proportions of abnormal images and normal images.Experimental results show that the proposed method can reduce the mutual interference between abnormal medical images and normal medical images and achieve complex medical image clustering.