Research On Edge Detection And Segmentation Methods Of Cardiac Magnetic Resonance Images

Magnetic resonance imaging is one kind of non-interventional imaging modality, magnetic resonance image plays a supporting role in the doctor's diagnosis. The analysis of magnetic resonance image using computer can improve the efficiency and accuracy of the doctor's diagnosis, and is of great significance to disease prevention and early treatment. The paper has studied the edge detection and segmentation methods of cardiac magnetic resonance image. The main contributions are as follows:(1) The edge detection algorithm based on gravitation is improved. Although the edge detection algorithm based on gravitation can accurately find the target edges, while some other edges may be missing and the algorithm is very sensitive to the noise. Motivated by the Convolutional Virtual Electric field and we have improved the edge detection algorithm based on gravitation. The improved algorithm can effectively suppress noises, and keep more edges. Taking into account the single impulse response criterion of Canny edge detector, the method of non-maxima suppression is employed to refine edge.(2) The R-snake model is used to segment the endocardium and epicardium of the left ventricle in cardiac magnetic resonance images(MRI). The Snake model uses a set of discrete points which are driven by internal energy and external energy to approach a target contour, and the calculation speed of the model is slow. The R-snake model approaches the target contour using fewer control points to control a continuous and smooth Gaussian curve, and does not need the internal energy. The calculation speed of the R-snake model is faster. In this paper, we use R-snake model with constraints to segment the cardiac MRI and achieve automatic segmentation of the endocardium and epicardium of the left ventricle.