Level set formulation based on edge and region information with application to accurate lesion segmentation of brain magnetic resonance images (Q2317855)

One of the main challenges in the automatic segmentation of brain MRI images is intensity inhomogeneity. This paper proposes a new algorithm which can handle this problem. The algorithm is based on minimizing an energy functional computed from two level set functions, which can segment an image into four regions. This is enough to distinguish white and gray matter and cerebrospinal fluid. The energy functional includes an edge detection term, two atlas functions, and a weight function which balances local and global terms dynamically. The functional is minimized by a split Bregman algorithm. The model is applied to 35 different brain MRI images with lesions, and compared to other methods. Experimental results demonstrate accuracy, efficiency and robustness of the proposed algorithm.