| At present,many scholars have proposed a lot of segmentation algorithms for the internal structure of the heart,and most of these algorithms focus on the segmentation of the heart specific atrial and ventricular structures.However,the role of the heart in the actual medical research is more than that,and it is necessary to obtain a complete cardiac structure for the application of early auxiliary diagnosis,interventional therapy and cardiac surgery navigation.Therefore,it is of greatly clinical value to accurately extract the complete region and edge of the heart and to provide a predictive model to assist the clinician in the early diagnosis and treatment of patients.The main contents of this thesis include:(1)Research on improved algorithm of fuzzy C mean based on neighborhood correlation information and kernel function.The fuzzy C means algorithm for image segmentation of gray value changes were fully considered,and the related information of neighborhood space relatively is lack of consideration,the noise is extremely sensitive,so the effect of access on the target area of interest in medical images is not very ideal.The proposed algorithm is based on the neighborhood information of pixels and the fuzzy C mean algorithm of kernel function.In this paper,the improved method can be used to segment the heart beat cycle image.(2)An improved algorithm for active contour model based on entropy and edge guidance function.When the active contour model is used to segment the heart image,it is necessary to change some feature information which is similar to the target to the user preset energy information and then put it into the active contour framework.According to the image of the objective and real data,the entropy method can completely correct grasp the energy parameters of the contour model,which has a good objectivity.By introducing the edge leading function,the evolution of the curve can be better grasped so that it can be more close to the contour of the target.(3)The prediction model of heart.The so-called heart prediction model refers to dynamically observe the periodic motion of heart through analysis of cardiac magnetic resonance images,which sums up the movement trend of the heart beating period,and gives an estimate of cardiac image sequences. |