Segmentation Of Tumor And Edema Based On K-Mean Clustering And Hierarchical Centroid Shape Descriptor

This study investigates a novel technique of tissues segmentation of high-grade (HG) glioma. Segmentation of tumor and edema for treatment planning is crucial. Anisotropic diffusion filter removes the noise and preserves the tumor tissues in MRI images. K-mean clustering algorithm clusters the brain tissues in normal and tumor tissues. The healthy tissues surround tumor tissues. Hierarchical centroid Shape descriptor select the tumor tissues and discard the other healthy tissues. The features of multimodality MRI images are fused to segment the tumor and edema. The experiment was carried out on the 200 T1, T1c, T2 and Flair MRI images of 10 high-grade glioma patients. The quantitative evaluation of experiment was carried out over publically available synthetic images from the BRATS2012 database.