Data Augmentation using Generative Adversarial Networks for Pneumonia classification in chest X-rays

In medical images, data augmentation is essentially important for accurate classification of images especially when available data is limited. This paper proposes a noble data augmentation method of generating synthetic chest X-ray images of patients with Pneumonia using Generative Adversarial Networks (GANs). The proposed model first uses conventional data augmentation techniques along with GANs to generate more training samples. A specific implementation of GANs allows us to produce unprecedented Chest X-Ray images of patients suffering from Pneumonia. The generated samples are then used to train a DCNN model to classify chest X-Ray images. The classifier significantly improves its accuracy after the introduction of synthetic data produced by the GAN model.