SynthXNet: Adversarial Learning for Data Augmentation in CoVID Severity Classification from X-Rays

For efficient patient care and management, the correct and prompt assessment of CoVID severity using X-Ray images is crucial. Existing approaches for classifying CoVID severity from X-ray images have shortcomings in precision, accuracy, recall, and latency; as a result, more reliable and effective ways must be developed. This study introduces "SynthXNet," a unique approach that uses adversarial learning and data augmentation to improve the performance of CoVID severity classification, in response to such demands. A substantial collection of X-Ray images is acquired, covering cases of CoVID patients with differing degrees of severity, spanning different age groups, genders, and ethnicities, to address the dearth of diverse and representative datasets & samples. The Generative Adversarial Network (GAN) architecture, which consists of a generator and discriminator network, is the central innovation of this work. To create artificial images that closely mimic actual X-Rays of CoVID patients, the GAN is trained on the X-Ray dataset that has been gathered. In an adversarial training procedure, the generator and discriminator teach each other to produce realistic images while the discriminator teaches each other to discern between actual and fake X-Rays. A deep learning model, specifically a Convolutional Neural Network (CNN), used for CoVID severity classification uses the synthetic X-Ray images produced by the GAN as extra training data. In comparison to recently proposed methods, the model’s precision increases by 8.5%, accuracy by 9.4%, and recall by 8.3% when the dataset is supplemented with these synthetic images. The suggested method also speeds up CoVID severity classification by 4.9%, making it more effective for real-time scenarios. In order to overcome the shortcomings of existing approaches, "SynthXNet" presents a potent solution to CoVID severity classification from X-Ray images by utilizing adversarial learning and data augmentation. Its potential to improve patient care and management in the context of CoVID severity evaluation is indicated by the results, which show significant increases in performance indicators.