The Research Of Human Brain MRI Segmentation Methods Based On Gaussain Mixture Model,Markov Random Field And Fuzzy Clustering

Medical image segmentation is an important prerequisite for clinical medical diagnosis and is also an important part of medical image analysis and application.For medical images,the different parts of the image are often used to analyze different structures,tissues and lesions.For medical diagnosis,doctors are only interested in certain parts of the image which is what we need to divide.At present,in the research of medical image segmentation,there are a lot of methods for the segmentation of human brain MRI.Segmentation methods based on threshold,region,edge or probability statistics are used commonly,but there is no one that can be adopted to various targets effectively.Segmentation methods which are really used in clinic are little.The manual segmentation is still a main segmentation method.However,the manual segmentation is a time-consuming and difficult task,its segmentation result depends largely on the doctor's professional knowledge and the segmentation results are very different.So the study of automation human brain images segmentation methods has been a popular research direction in the field of image processing.The human brain image has an irreplaceable role for the diagnosis of human brain diseases.Such as brain disease segmentation images of brain tumors,cerebral infarction,Parkinson syndrome can help doctors to perform fast and accurate diagnosis.However,the imaging principle of MRI makes the image show a certain ambiguity,which makes the boundary of each soft tissue unclear and discontinuous.So the accurate segmentation of the human brain MRI is greatly difficult.At the same time,the imaging process of acquisition,transmission,compression and decoding,thermal and electrical noise,RF coil and magnetic field effect of the inhomogeneity and partial volume effect makes the human brain MRI shows fuzzy and noisy,the noise will seriously affect the image segmentation and increase the segmentation difficulties.Based on the above reasons,this paper hopes to abtain a robust,anti noise and generalization human brain MRI automatic segmentation method by researching different segmentation methods.The main contents of this paper are as follows:(1)Due to the excellent spatial correlation of Markov random field that can be good for human brain MRI texture and edge and the gray distribution of human brain MRI showed Gaussian distribution characteristics,this paper adopts Gaussian function to establish the Markov random field model which can fit well with the distribution of human brain MRI;Since the Markov random field is sensitive to noise,in this paper,a filtering method which is suitable for Markov random field is proposed;Due to the high calculation cost of simulated annealing algorithm when solving the Markov random field,the stable and unstable points of simulated annealing algorithm is used to solve the Markov random field and can effectively reduce the computational cost and improve the timeliness of calculation.(2)The traditional clustering segmentation method is usually divided pixel only based on image gray value similarity,so the effect is not ideal for noisy or blurred edge human brain MRI,this paper proposes an improved FCM algorithm which can effectively reduce the influence of noise;At the same time,the improved FCM algorithm as the initial method can play the good performance to generate excellent initial parameters,it effectively improve the segmentation effect.(3)For the disadvantage of poor anti noise performance and robustness of traditional Gaussian mixture model,this paper presents a brain MR images segmentation method based on the hidden Gaussian mixture model.Due to neglect of the spatial information and the segmentation results distribution,the traditional Gaussian mixture model is incomplete.In response to these shortcomings,in this paper,The probability density function of the segmentation results which is regarded as the hidden data is introduced into the Gaussian mixture model and a nonlinear weighted hidden Gaussian mixture model is established.Meanwhile,the Gaussian weighted exponent which contains spatial information and the smoothing factor is introduced.And EM algorithm and Newton iteration method are used to calculate the class mean and variance,and the smoothing factor.Finally,the segmentation results are obtained according to the maximum a posteriori criterion.