Cardiac Mri Segmentation Based On Deformable Models

Research on cardiac MR image (MRI) analysis is a significant topic in the field of medical image for many years. Currently, this research chiefly focuses on motion and strain analysis of the left ventricle (LV). Consequently, it is principal to precisely extract the LV from cardiac MRI.Image segmentation using deformable models offers a novel method for effectively extracting the region of interest (ROI) in the images, and it has been an important tool of the medical segmentation. Deformable models consist of the parametric active contour model and geometric active contour model. In this paper, the two models are analyzed detailedly. Moreover, some research work on their application in the segmentation of cardiac MRI are presented.When parametric active contour model (Snake) is applied to segment the image, an initial contour must be set near the boundary of ROI and the model is difficult to segment deeply concave regions accurately. Therefore, Cohen added a balloon power in his model, which can enlarge the capture rage. But it also depends on the initial contour. Here, on the basis of analyzing the Ant Colony Algorithm (ACA), an improved parametric active contour model is proposed. First, gray statistical characteristics are used to automatically initialize Snake. Then, a centripetal energy is added to Snake model in evolving process so as to move into the concave regions. Finally, ACA optimizes the result and make it converge to the whole optimum. The proposed model has no special requirement to the location and shape of the initial contour. The encouraging effectiveness of the model is demonstrated with the experiments on the cardiac MRI.Snake model is difficult to change the topology and extend to higher space. However, the geometric model based on Level Set can resolve these problems by embedding lower space into higher space. Geometric model can change the topology freely, but its computational magnitude is too large. Moreover, Due to cardiac deformation and blood flowing, weak edges, local gradient maximum regions and artefacts are found in the MR images. However, active contour models usually use information of gradient and is difficult to obtain the ideal results. Here, information of ROI is applied in pressure force so as to reduce influence of the Noise and weak edges. Then using the mathematical relation of parametric and geometric models, the active contour model is made geometric so as to have the ability of changing topological structures. The experiment results demonstrate that this algorithm can segment MR images effectively.