Chest X-Ray Images Generation Using GAN
Student
Iris Yan
Advisor
Daniel Boley
Abstract
During the pandemic, many X-ray images are needed to train a classification model. However, because of the privacy issue, medical image datasets with labels are in a small size and highly imbalanced in the class distribution. This project implements a GAN to generate chest X-ray images for data augmentation and evaluates these generated images. A Deep Convolutional GAN (DCGAN) model is trained for each class, and its generator is used to produce fake X-ray images in that class. And then, qualitative evaluation and quantitative metrics are applied to measure the quality of the generated images.