Medical Image Clustering Segmentation Based On Intuitionistic Fuzzy Set

With the development at full speed of modern science and technology,some advanced imaging technology such like computed tomography,magnetic resonance imaging,ultrasonic imaging, positron emission computed tomography and so on,is widely used in clinical diagnosis and treatment,which become the important basis for doctors to analyze disease.At the same time,it provides scientific basis for biomedical research.However,due to the limitations of imaging equipment and environmental influences,medical image is often blerred and contains noise.In order to provide doctors with more complete and accurate information,segmenting the interested parts accurately and noiselessly have been popular subjects for many scholars.Medical image segmentation then become a necessary prerequisite for measuring a particular organization,extracting a pathologic region or reconstructing an organ.For the fuzzy characteristics of the medical images,scholars put forward using the fuzzy theory to segment them.The concept of precise set description is either this or that.Fuzzy set can describe situation which is both this and that.Intuitionistic fuzzy set has added a new attribute parameters,hesitancy degree.Then it can also describe the fuzzy concept of neither this nor that,which express the extent of the neutral.Compared with the fuzzy set,intuitionistic fuzzy set can better and more accurately express the fuzzy information.Therefore,this paper mainly studied the medical image segmentation algorithm based on intuitionistic fuzzy set.At First,the background and significance of medical image segmentation is presented.And then the concept of fuzzy set and intuitionistic fuzzy set is introduced.By amending distance measurement opetaor and joining some constraint items to the membership function,some improved intuitionistic fuzzy clustering algorithms are proposed.A quick intuitionistic fuzzy clustering algorithm and its modified form are also proposed.All these algorithm are applied to medical images to verify its effectiveness. The main work is as follows:(i) Using the kernel function to replace the Euclidean distance measuring formula.First ofall,the function of kernel function is discussed and several kinds of common type of kernel function are introduced. Secondly,radial basis kernel function,gauss kernel function and hyperbolic tangent kernel function are used to improve the traditional intuitionistic fuzzy clustering algorithm.By comparing with the corresponding fuzzy clustering algorithm to cluster the medical image containing noise,the superiority of intuitionistic fuzzy set is verified.To produce intuitionistic fuzzy set,the Sugeno type and Yager type intuitionistic fuzzy generator are used.The advantages and disadvantages of these two approaches are introducesd from two aspects of treatment time and the final results.Finally,on the basis of all these,a new intuitionistic fuzzy clustering algorithm based on kernel function using neighborhood information and sample variance is proposed.Algorithm is verified through the experiment in the aspect of noise suppression has improved further.(ii) The space operator constructed by the local spatial information is studied,which used to adjust membership function to improve the traditional intuitionistic fuzzy clustering algorithm.At first,a local intuitionistic fuzzy clustering prototype put forward by B.K.Tripathy is introduced.The constructure of the spatial function and the correction of membership function are minutely limned. The superiority of the algorithm with respect to the fuzzy clustering is verified.Then,a new kind of space function constructed by Gauss space kernel function is proposed.The accuracy of the clustering results of the proposed algorithm is verified from two aspects of the visual results and the quantitative evaluation parameters.(iii) Aiming at the shortcomings of long processing time of the clustering algorithm based on intuitionistic fuzzy set,a fast intuitionistic fuzzy clustering algorithm based on the fast fuzzy clustering algorithm is proposed.Compared with the traditional intuitionistic fuzzy clustering,it has improved greatly in time complexity.On the basis of this fast intuitionistic fuzzy clustering algorithm,an improved fast intuitionistic fuzzy clustering algorithm is proposed by using the relative position information and the gray level information of neighboring pixels.These contribute to the algorithm acheiving balance in three aspects:noise suppression, retention of details and processing time.