AI-Driven Synthetic Data Generation Using Conditional GANs to Mitigate Class Imbalance in COVID-19 X-Ray Imaging

The current pandemic of the new coronavirus disease has emphasized the need for fast and specific diagnostic approaches for medical image analysis. Conventional interpretation of X-rays requires a lot of work and can produce wrong results hence research towards automated procedures. To improve the COVID-19 diagnosis and analysis task due to the shortage of data, we present a new method using Conditional Generative Adversarial Networks (CGAN) to generate artificial chest x-ray images. The synthetic images are employed to enhance the performance of a Convolutional Neural Network (CNN) for classification to distinguish between COVID-19, pneumonia, and normal chest X-rays. Depending on the classification in all of the classes, the presented model of 99.45 % recognition accuracy is state of the art, while that for COVID-19 is 99.50%; that for normal is 99.40%; and that for pneumonia is 99.45%. This demonstrates the readiness of the improved GANs data in medical imaging which ensures the scalability of assistive diagnosis solutions. The novelty of this study does not merely lie in the confirmation of GANs as a reliable method of data augmentation but in the expansion of the possibility of GANs in diagnosis in the field of healthcare.