Classification of Tissue Types in Histology Images Using Graph Convolutional Networks

This article uses Graphic Neural Network (GNN) models on histology images to classify tissue to find phenotypes. The majority of tissue phenotyping approaches are confined to tumor and stroma classification and necessitate a significant number of histology images. In this study, Graphics Convolutional Network (GCN) is used on the CRC Tissue Phenotyping dataset, which consists of seven tissue phenotypes, namely Benign, Complex Stroma, Debris, Inflammatory, Muscle, Stroma, and Tumor. First, the input images are converted into superpixels using the SLIC algorithm and the region neighborhood graphs (RAGs), where each superpixel is a node, and the edges connect neighboring superpixels to each other are converted. Finally, graphic classification is performed on the graphic data set using GCN.