Which Data Subset Should Be Augmented For Deep Learning? A Simulation Study Using Urothelial Cell Carcinoma Histopathology Images

In this article, researchers evaluated the ability of four pre-trained convolutional neural networks (CNNs) to discriminate between urothelial cell carcinoma (UCC) and inflammation of the urinary bladder. This simple classification task was used as a benchmark for the experiments to reduce the confounding factors to a minimum. The overall workflow of the methods comprised dataset building, dataset preprocessing, model building, model testing, and statistical analysis. The dataset source was the Cancer Genome Atlas (TCGA).