Tumor segmentation from small animal PET using region growing based on gradient magnitude

In this paper, we present a region growing algorithm based on gradient magnitude for tumor segmentation from small animal PET images. The segmentation of PET images with low spatial resolution and high variations of intensity is more difficult than images with high resolution. Especially, it is not easy to detect the boundary between region of interest and abutting regions with similar intensity. We propose a region growing algorithm to extract tumor from adjacent regions with similar intensity values using gradient magnitude difference. It can accurately extract the tumor and reduce the time for manual post-processing due to over-segmentation result from ordinary region growing with intensity criteria.