Automated Brain Tumor segmentation using novel feature point detector and seeded region growing

In this paper, we propose a methodology for fully automated Brain Tumor segmentation from T1 weighted contrast enhanced Magnetic Resonance Images. A novel algorithm has been designed to extract the visually significant feature points. Feature points relating to Tumor are then identified and extracted as seeds for further region growing. Feature points are obtained by fusion of wavelet methods and image edge maps. Robustness of feature points to geometrical transformations and scaling have been shown. Our method gives a sparse representation of the information (region of interest) in the medical image and thereby vastly improves upon the computational speed for tumor segmentation results.