An interpretable COVID-19 chest X-ray classification through transfer learning and discriminative localisation-based deep learning techniques. (7th December 2022)

• Record Type: Journal Article
• Authors: Yamathi, Lavanya; Rani, K. Sandhya
• Is Part Of: International journal of smart grid and green communications. Volume 2:Number 2(2022)
• Page Start: 115
• Page End: 131
• Publication Date: 2022-12-07
• Subjects: COVID-19 -- discriminative localisation -- deep learning -- transfer learning -- convolutional neural network -- CNN; Smart power grids -- Computer networks -- Periodicals; Smart power grids -- Periodicals; Wireless communication systems -- Security measures -- Periodicals; Wireless communication systems -- Energy consumption -- Periodicals; 333.7911028546
• Languages: English
• ISSNs: 2052-2002

Abstract

The present time is marked by the upsurge of coronavirus (COVID-19) pandemic, which persists and has catastrophic consequences on global health and well-being. In addition to RT-PCR test, CT scan and chest X-ray have become essential in detecting and treating COVID-19 patients. Several deep learning frameworks have been put forward in recent times for the COVID-19 chest X-ray classification. Therefore, to overcome the challenges of data scarcity and lack of interpretability, also to increase the performance of COVID-19 chest X-ray classification, a first of its kind of model is proposed in which transfer learning (TL) and discriminative localisation (DL) are successfully adopted. To verify the superior classification performance of the TL-DL-based C-19CXC model, a set of experiments are conducted on widely used eight pre-trained models like MobileNet, SqueezeNet, etc., on publicly available large datasets. The MobileNet based model outcome accuracy is 98.73% followed by Xception-based framework with 98.34% accuracy.