Research On The Active Contour Model And Its Application To Medical Image Segmentation

Medical image segmentation is the base of medical image processing and analysis, many researchers pay close attention to it in recent years. The purpose of the medical image segmentation is to discriminate different regions in image, and make the regions close to the human anatomy as far as possible. The segmentation result can be an evidence of clinical diagnosis and pathology research. So research on the medical image segmentation has important significance to science and worthiness in practical applications.In this dissertation, we survey medical image segmentation technology and analyze active contour model (ACM). A new ACM based on statistical shape is presented, and this method segment the noise image, broke object well. The main aspects of research include:First, two segmentation algorithms of ACM are analyzed in this dissertation; the advantage and disadvantage of those method are presented by experimental.Second, a new hybrid method base on geodesic active contour model and statistical shape energy, named as statistical shape active contour model (SS-ACM), is proposed, and it is applied to defect lung segmentation. In the model, statistical shape energy can guide and limit the evolution curve to the target. The experiments show, this method can get good results when the target is defect or polluted by noise.Third, for the application of lung segmentation in chest CT image, a semi-automatic segmentation is proposed. This method includes regional growth lung segmentation and lung area repair. Chest CT images are segmented by regional growth, and the seed points are selected automatic for the similarity between image sequences. Then, for broken lung areas, firstly, the rolling ball method is employed for repairing; if the repair result is unacceptable, the SS-ACM is applied to repair the lungs.Finally, a system is implemented for chest CT images segmentation.At the end of this dissertation, experiments results are carried out to evaluate the performance of the methods above. It supports the feasibility of these methods.