The Effectiveness of Image Augmentation in Breast Cancer Type Classification Using Deep Learning

The past few decades have witnessed a surge in applying machine learning methods in the medical fields to assist in diagnosing various medical conditions. Among all the methods, the convolutional neural network (CNN) has archived the best performance, and researchers have promised many CNN variants to edge the performance in different specialized areas. It has been shown that image augmentation can help improve the CNNs' performance in the image classification task. In this study, we explore how image augmentation methods may help CNN models improve classification performance in predicting biopsy slides benign (not dangerous) or malignant (uncontrolled and may cause death) from breast cancer tumor tissue. We perform a qualitative study on the effectiveness of some basic image augmentation methods.