Study On Medical Image Segmentation Method Based On Parametric Active Contour Models

Medical image segmentation is a fundamental technique of medical image processing. Traditional image segmentation methods have a lot of shortcomings and hardly meet the demands of the complex medical image segmentation. Parametric active contour models (or snake models) are represented explicitly as polynomials or splines. Given an initial contour, the evolution of a parametric active contour model is driven by external forces derived from the image while the shape of the contour is maintained by the internal forces. Parametric active contour models are effectively used for medical image segmentation.A survey on parametric active contour models is given in this thesis. The relative theories and methods for the improved snake models and the applications for image segmentation are studied. Then the traditional snake and the GVF snake are studied and some experimental results are shown.According to the characteristic of medical images, a novel external force for snakes named dynamic directional convolution vector field is proposed in the thesis. The method is validated on both phantom and clinical images. Experimental results show that the method prevents the disturbance easily, such as strong edges and edges near the desired boundary. Additionally, the DDCVF snake has larger capture ranges and better robustness to initialization and it has better segmentation results for complex medical images.Due to the diversity of medical images and the existence of artifacts and noise, when the initial contour is not suitable, the DDCVF snake may not get the exact and stable segmentation results. The author then combines wavelet transform with the DDCVF snake. Experimental results show that the capture range of snake is effectively enlarged and the method further reduces the influences of noise and initialization. It is more applicable to the field of medical image segmentation.Finally, the summary and the prospect on this study are given.