Active contour model combining local and global information dynamically with application to segment brain MR images

Abstract

With the rapid development of medical imaging technology, the image segmentation has a special significance in medical applications. It's known that intensity inhomogeneity is one of the important features of magnetic resonance (MR) images, which presents a quite challenge in MRI segmentation. In this paper the authors apply the split Bregman method for minimization of the improved active contour model combining local and global information dynamically to segment brain MR images. The authors have proved this model can segment synthetic and real images with intensity inhomogeneity. Numerical results show the accuracy and efficiency of this model. Besides, this model is also robust to noise. That is exactly the reason why the authors apply this model to segment brain MR images. The authors present this model in a multi-phase formulation and use it to segment brain MR images with multiple regions adjacent to each other. Then the authors have tested this proposed model with many brain MR images. Finally, comparisons with other models and experimental results have demonstrated the efficiency and accuracy of this method.