Automatic left ventricle segmentation in cardiac MRI via level set and fuzzy C-means

Magnetic resonance imaging (MRI) has become an important assistant for clinical diagnosis of cardiac diseases which can not only observe the morphological structure of the heart, but also estimate the global and local function of myocardium. It is necessary to segment the left ventricle (LV) for the quantitative analysis of the global and regional cardiac function. However, cardiac MR images are usually intensity inhomogeneity, which results in a considerable challenge in left ventricle segmentation. In this research, we presented a synthetically automatic LV segmentation model on basis of modified level set and fuzzy C-means. We used level set method to delineate the endocardium and estimated the bias field which was used to decrease the intensity inhomogeneity of cardiac image. In addition, the fuzzy C-means algorithm and morphologic segmentation were applied in the corrected MR image to segment the epicardium. For the algorithm evaluation, we tested the short axis cardiac cine MR images published by MICCAI. The experiment results showed that our method obtained a good performance for both the endocardium and the epicardium segmentation. And, it was more effective to delineate epicardium in the corrected image than the original image.