Brain Tumor Segmentation Research Based On Active Contour Models In Magnetic Resonance Images

Computer related techniques have been broadly applied in medical fields in recent years, which accelerates the coming of the age of digital medicine. The image processing is the key technology in this field. Especially image segmentation is a hard and basic work for many important applications, such as3D reconstruction, Image Guided Surgery, Computer-aided diagnosis and so on. The task for this paper is to find a reasonable way to finish a good work for the MRI brain tumor segmentation.Region-scalable fitting (RSF) energy model is a region-based active contour model defined with intensity information in local regions at a controllable scale. Incorporated into a Level Set formulation it can realize the perfect segmentation for images with intensity in homogeneity and noise. This feature is fitting for tumor segmentation in MRI. However, it has several complex parameters. And the complex intensity of MRI always forces the model to make wrong convergence. This paper improves the model with a changed Level Set formulation. It overcomes the limitation of RSF to make it more fit for brain tumor segmentation in MRI. To deal with some more complex condition we also combine the mean-shift method with our model in special condition. The application for the real image proves the improved model can realize fast, accurate, robust segmentation for the tumor tissue in MRI.