Right Ventricle Segmentation Based On Cardiac Magnetic Resonance Images

Cardiovascular disease is the leading cause of human health with the high fatality rate and the increasing morbidity.It is a common method to diagnose the illness by the Cardiac magnetic resonance(CMR)images.However,it is time-consuming,laborious and error-prone to segment CMR manually,and automatic segmentation by medical image segmentation technology is very necessary.Compared to right ventricle(RV),left ventricle is regular shape and its segmentation is easier.There are a lot of research results on it.However,medical research shows that the evaluation of right ventricular is very important for some heart disease diagnosis,while there is a little research on its automatic segmentation.So it makes much sense to research on automatic RV segmentation.This paper is a study of RV segmentation based on short CMR images.The mainly works are described as follows:1.We try to segment RV with the traditional segmentation methods,and point out the limitations of these methods and their inadaptability of the RV segmentation.2.Caused by intensity inhomogeneity and low contrast between the myocardium and cavity of the CMR images,it is necessary to preprocess the images.Based on the analysis of common filtering methods,we design a CMR filtering method based on anisotropic diffusion to remove the biased field,and lay a foundation to complete the follow-up accurate segmentation.3.We propose a new method based on a combination of simplified pulse coupled neural network(SPCNN)and morphological operation.Among them,the endocardium segmentation takes advantage of the synchronous ignition characteristic of the SPCNN to distinguish between the images’ regional information and edge information,and obtain an accuracy,robust result while the operation is simple.Caused by the shape similarity to the endocardium,the morphological operation,which is simple and effective,is used to obtain RV epicardium.In this paper,the proposed method is test on the database which is provided by Medical Image Computing and Computer Assisted Intervention(MICCAI)2012.The subjective and objective evaluation on image preprocess,segmentation results show the good performance,and prove the effectiveness of the proposed method.