Region Labeling Based Brain Tumor Segmentation from MR Images

Brain tumor due to their increasing rate and high uncertainty has become a curse for mankind. For their effective diagnosis, automatic systems called Computed Aided Diagnosis (CAD) systems have developed that help in tumor analysis without manual interference. However, due to high variability in tumors, their segmentation from MR images is a challenging task. This paper proposes an improved tumor segmentation methodology that is an extension to simple thresholding technique. In this method, different regions of the binary images are labeled and are segregated on the basis of solidity and area. Then the region having solidity around 50% and maximum area is extracted as tumor. This segmented region is further dilated to include edema tissues surrounding the tumor. The performance of the methodology is justified by the results obtained in which only the tumorous region has been extracted indicating successful segmentation without inclusion of other brain tissues.