Research On MR Image Segmentation Algorithm For Infant Brain With Spatial Information FCM And Improved Level Set

Infantile period is the fastest growing period of human brain.During this period,the brain is very strong,but it is easy to get all kinds of brain diseases,but in this period infants and young children are unable to express their condition,so a good medical diagnosis method is very important.Magnetic resonance imaging can detect human internal lesions without generating human trauma,and generate high-resolution images for doctors to diagnose.Therefore,MRI technology has been widely used in the diagnosis of infant brain diseases.Due to the influence of device conditions and other factors,infant brain MR images can cause uneven gray level,noise and partial volume effect.So the key to get good images is the pre-processing of images.The main methods of the image segmentation algorithm,the classical algorithm used is the fuzzy clustering algorithm,but the algorithm is highly sensitive to noise,so many researchers have made improvements to this algorithm,this thesis summarizes some improvement based fuzzy C means algorithm,according to the comparison of two representative algorithms,to improve it,and some study on the combination of fuzzy clustering and level set method.The main contents of this thesis are as follows:(1)Aiming at the problem of infant brain MR image noise,fuzzy C means algorithm using adaptive space,this method can calculate the adaptive parameters of the whole image,and its application in segmentation,on the basis of joining the nuclear function and non-local method,the formation of a new algorithm,this algorithm takes into account the global and local the spatial information,and this information will be considered in the objective function.The experiment shows that the improved algorithm has better effect on the noise image segmentation.(2)Segmentation based on fuzzy C mean algorithm of pixel accuracy problems,according to the membership degree of fuzzy C means to improve the degree of membership to join a new definition on the basis of the original,this is no longer limited to various types of membership and 1,expanded the scope of the membership,so that the relationship between class and class the introduction of objective function.On this basis,the kernel function and non-local method are added to increase the neighborhood information capacity.Good segmentation results are obtained in the experiment.(3)In view of the problem of poor edge segmentation algorithm,fuzzy clustering and level set combined with fuzzy clustering method to get the approximate position of image clustering,membership function,and then the membership function is applied to the new level set method,iterative level set,a new method of the level set iteration speed accelerated by fuzzy clustering,and enhance the accuracy of the segmentation.