ABSTRACT
This chapter summarizes the cardiac magnetic resonance (CMR) image segmentation and the challenges and the level set functions, some commonly used segmentation methods. It provides the formulation and the numerical implementation of the stochastic image segmentation. The main idea of CMR image segmentation is to identify the boundaries of cardiac chambers and separate them from the background. As compared to image segmentation with fixed pixel values, probability density functions are describing the distribution of pixel values and to quantify the effect of uncertainty on image segmentation. The epicardium is in the middle of the myocardium and the surrounding tissues such as fat and lung, which may have different pixel value profiles and may show lower contrast as compared to myocardium. A generalized polynomial chaos expansion is combined with a level set function to evolve the boundary in a stochastic way. The boundaries of endocardium and epicardium identified from an image segmentation algorithm can be used to quantify the heart function.
