Automated classification algorithms in the medical field are especially resourceful for tissue analysis, diagnosis, and disease detection. Classification decisions are made based on the detection outcome, with the goal of matching the ground truth set by medical professionals. While algorithms in the past have employed cell-based analysis for detection of tumor types in specific regions of interest, none have focused on classifying whole brain tumor samples according to tumor grade. The algorithm proposed in this paper distinguishes between low-grade glioma (LGG) and glio-blastoma multiforme (GBM) grades of histopathological brain tumor tissue. The image samples used for the development and testing of our approach were provided by The Cancer Genome Atlas (TCGA), a national program responsible for the collection of different tumor types for research purposes. We have also worked in collaboration with the MD Anderson Cancer Center in obtaining ground truth labels for the tumor samples. Our goal is to match this diagnosis as closely as possible.
