Accurate Segmentation Of The Right Ventricles In Magnetic Resonance Images Based On K-SVD Data Dictionary Learning

Cardiovascular disease is a serious threat to human health and especially the common disease of the middle and old aged people over 50 years old,even today,with the most advanced and perpfect treatment,there are still 50%of survivors can not fully take care of themselves,so how to quickly and accurately diagnose and effectively treat cardiovascular disease has becoming more and more concerned about the problem.Research results have been confirmed,and left ventricular abnormalities in the diagnosis of cardiovascular disease such as heart failure and ischemic heart disease and other symptoms of the prediction and diagnosis of the same,the detection and diagnosis of the morphology and function of the right ventricle can play a very good role in the detection and diagnosis of the right ventricular morphology and function.The advantages of magnetic resonance imaging in the clinical diagnosis of heart disease has been widely recognized,accurate segmentation of cardiac magnetic resonance images has become a research focus in the field of medical image processing.However,because the heart cavity complex geometry and and the characteristics of tissue around the special anatomical structure,high variability,unique thin wall,the boundary is not obvious and the low contrast,whether the right ventricle can be successfully segmented into the key factors of success or failure of the heart.Firstly,in this paper,based on the study of the anatomical structure of the heart,the threshold method,the regional growth,K-Means and other segmentation algorithm for the right ventricle segmentation on the MATLAB platform.The experiments show,this series of segmentation has a lot of limitations,such as,K-Means is divided into too many areas,can't handle target area and background area very well,so the right ventricle segmentation is not accurate enough.Therefore,this paper proposes a segmentation algorithm based on K-SVD dictionary learning.The segmentation algorithm based on K-SVD is used to train the dictionary atoms in a large amount of data,the segmentation problem is transformed into a linear sparse representation of the training dictionary,extracting the corresponding features of the target region from the segmented images and the target image,and then according to the training data obtained after the training of the dictionary,then the sparse coefficients obtained from the OMP to the input images are obtained in the training dictionary,according to the product of the dictionary and coefficients to reconstruct the pixel features,re classification and marking according to the error of the actual feature and reconstruction feature,then change the pixels around the same pixel to the same pixel to get the final segmentation result.The premise of the algorithm is the use of the training image set and the target image in a sense is quite similar,So how to select the training image set is particularly important.At the end of this paper,the data provided in the MICCAI right ventricular segmentation challenge match is supported,according to the above method,the right ventricle segmentation,and the segmentation results are analyzed,it is found that the segmentation method based on K-SVD training dictionary is more accurate,which proved to be effective method in this paper.