Generalization of Deep Neural Networks for Chest Pathology Classification in X-Rays Using Generative Adversarial Networks

Medical datasets are often highly imbalanced with over-representation of common medical problems and a paucity of data from rare conditions. We propose simulation of pathology in images to overcome the above limitations. Using chest X-rays as a model medical image, we implement a generative adversarial network (GAN) to create artificial images based upon a modest sized labeled dataset. We employ a combination of real and artificial images to train a deep convolutional neural network (DCNN) to detect pathology across five classes of chest X-rays. Furthermore, we demonstrate that augmenting the original imbalanced dataset with GAN generated images improves performance of chest pathology classification using the proposed DCNN in comparison to the same DCNN trained with the original dataset alone. This improved performance is largely attributed to balancing of the dataset using GAN generated images, where image classes that are lacking in example images are preferentially augmented.

1. Introduction

In the medical domain, preservation of patient privacy is paramount, and hence access to data is often intrinsically limited to research groups [1], [2]. Medical datasets, similar to financial [3] and genomics [4] datasets, are also very limited because they are often imbalanced [5]. Such imbalances in datasets potentially make the training of neural networks with equally high accuracy across classes technically challenging. Some medical problems are commonly encountered in hospital settings which leads to a substantial amount of data associated with them. However, rare conditions or syndromes such as Birt-Hogg-Dube syndrome are expected to have limited amounts of data in clinical databases [6].