Level-set segmentation of brain tumors in magnetic resonance images

Abstract

Three-dimensional segmentation is reliable approach to achieve an accurate estimate of the tumor volume. Among all possible methods for this purpose, the level set is a powerful tool which implicitly extracts the tumor surface. In this thesis, we propose two level-set based approaches for 3D tumor segmentation. The first approach introduces a threshold-based scheme in which the level set speed function is designed using a global threshold. This threshold is iteratively updated throughout the evolution process. In the second approach, one-class SVM (support vector machine) algorithm is integrated into the level set method to define an appropriate speed function for it. The SVM training is iteratively refined as the level set grows. These approaches are examined on 16 MR images and the experimental results confirm their effectiveness. Moreover, the comparison results among these approaches and also an existing region-competition based method indicate the superior performance of TLS.