Statistical Shape Modeling Based On Minimum Description Length Optimization And Segmenting In Medical Images

Along with the improvement of economics,health is one of the key topics in our modern society.Medical imaging provides a non-intervention overview of inner structures of our bodies and great help in diagnosis and cure.Combining fields like information science and medicine,medical image process and analysis is making another push in the progress of medicine digitally,automatically and intelligently.In this dissertation,research is done on Statistical Shape Modeling and Segmenting in intervertebral disc, vertebra and meniscus MRI.The main points are as follows:(1) we describe a minimum description length based optimization method for automatically building statiscal shape models from training set of example boundaries. A multi-scale parameterization on shapes allows the optimization on landmark correspondence in a coarse scale and a most convenient arc parameterization based landmark correspondence in a fine scale. This achieves a fast and accurate SSM building, which is the foundation on following image segmentation and quantitative analysis. In experiments, SSMs are built with vertebral body, intervertebral disc and meniscus shapes extracted from various MRIs respectively. It is testified that the models built with the proposed scheme is not only more repeatable and concise than model based on manually landmarking, but also more temporally efficient than model purely based on optimization.(2) Active Shape Model (ASM) is an efficient method of image segmenting. One key factor in building models is obtaining correspondent landmarks among the whole shape dataset. Traditional manual landmarking is temporally expensive, subjective,boring and prohibitively extensive in dimension. In this dissertation, a parameterization on shapes allows a Minimum Description Length (MDL) based optimization on landmark correspondence. Base on the point distribution models, we complete gray-level models building, which is the foundation on following image segmentation. The segmentation errors from the proposed method are comparable with or better than those from the manual modeling based segmentation.