Study On Medical Image Segmentation Based On Active Contour Model

Medical image mainly includes computerized tomography (CT), ultrasound image and magnetic resonance image, etc. It is very important for quantitive analysis, real-time monitoring and treatment scheduling, etc. When medical images are utilized to do medical analysis and diagnosis, it's usual to extract the specific organs or regions of interest (ROI) to do further analysis better, the process is image segmentation. Traditional image segmentation methods can not satisfy us because of medical image's low resolution, low contrast and intrinsic noises.Active contour model is newly used image segmentation and tracking methods by researchers in recent ten years. It researches the evolvement process of curve in dynamics angle and gets the ROI's continuous and close borders by computation. Practices prove that the method has great improvement than former image segmentation methods and adapts to medical image segmentation. But it is restrict to use in clinical practice in some degree because of problems itself. Based on the above facts, this thesis does all-around researches on active contour model to improve current algorithms or propose new methods for medical image segmentation. It's also a meaningful exploration for the segmentation of low SNR images.The thesis first gives an overview of the medical image segmentation methods. Then several classic active contour model algorithms and some algorithms that have been used in medical images are analyzed. Through the analysis the thesis finds that these methods have some shortage in some aspect. Then, the thesis carries on the work in following three aspects. Firstly, integrating with wavelet transform, the thesis improves the wavelet module maximum algorithm and increases precision and efficiency of the edge-detected algorithm, and then segments medical images integrating with gradient vector flow model. The method can overcome medical images' intrinsic noises and segment the objects' contour very well. Secondly, based on the newly presented electrostatic force active contour model, the thesis affirms the merits of the model's good antinoise, but at the same time it finds the model has the restriction of exterior forces. The thesis adopts adaptive weighting functions to extend traditional electrostatic active contour model, automatically sets weighting parameters based on image edges and dynamically controls the electrostatic external force of the model curve. The method can reduce the local minima in the external energy field, enlarge the range of the external force and improve the segmentation accuracy. Thirdly, the thesis studies the...