ABSTRACT
A two-stage segmentation algorithm is presented to solve the problems of inhomogeneity, weak edges and artifacts exhibited in the magnetic resonance imaging (MRI) images. First, the K-mean clustering algorithm is applied to classify the objects. Then, a speed function based on the clustering results is defined in order to search the rough boundary. Secondly, a speed function of the gradient intensity is constructed to locate the boundary accurately. Due to the lack of deformation information of the boundaries between MR slices, a deformable model is used to reconstruct the shape of the LV: a dynamic equation governing the surface deformation is given; from the slice data, external forces are constructed and elastic forces are provided with mean curvatures of the deformation surface. The level set method is applied to solve the dynamic equation for the LV shape. Experimental results demonstrate the effectiveness of the algorithm listed in the paper.
