The Effect of 3d-Mri Modalities Mixture in Glioma Delimitation

Abstract

today, image processing has become a very important issue in medical imaging field, which is constantly developing to facilitate the diagnosis of several diseases such as brain tumors, especially glioblastoma (GBM). The segmentation of glioblastoma tumors is an important early step in image analysis to characterize the tumor phenotypic features. This study describes a new approach for the detection and the delimitation of GBM using modalities mixture as a pre-processing step then Otsu multilevel thresholding and Neighborhood algorithm & maximum component. This proposed modalities mixture used three different MRI modalities which are Flair, T2 and T1. This approach has been tested on clinical database BRATS’2017. We report promising results. The Dice Similarity Coefficient metric for whole tumor was 0.88. the preprocessing step used increases the segmentation accuracy compared to the same technique without modalities mixture.