Computational Analysis for Candidate X-ray Images Using Generative Adversarial Network

Abstract

Medical image analysis is becoming a critical deep learning application based on machine learning by contributing to developing a more sustainable health system, which probably overpowers doctors’ workload drastically. With the advancement in deep learning technique, there is rise in the samples for training of diagnosis and treatment models. Generative Adversarial Networks (GANs) have sought attention in the field of medical image processing by their outstanding image generation capabilities and data generation without mapping the probability density function explicitly. GAN methods simulate the actual data distribution and reconstruct estimated accurate data. Medical images are available in less amounts, and the acquiring of medical image annotations is costly; therefore, generated data can solve the problem of data insufficiency or data imbalance. GANs are proven very useful in data augmentation and image translation. These qualities of GAN have fascinated researchers, and rapid adoption is noticed in the reconstruction, synthesis, segmentation, denoising detection, and classification of medical images. Finally, GAN models are extensively used for feature selection and extraction for medical image analysis and early diagnosis of diseases.