Grading of Brain Histopathology Images via Convolutional Neural Networks

Abstract

Grading is the process of determining the aggressiveness of tumors. Correct grading of histopathology images of the brain is very important for treatment planning. Pathologists examine the tissues with a microscope and decide which stage the brain tumors belong to. This process is time consuming and expert experience is important. With the advances in computer technology, computer aided image analysis on histopathological tissues has become possible. In this study, the AlexNet-based Convolutional Neural Network model was evaluated for automatic grading of astrocytomas using brain histopathology images. According to the simulation results, the proposed model reached 86.3% accuracy in the classification of all phases, while the low level histological images reached 98.38% accuracy.