COVIS-Health: A Deep Learning and Explainable AI-Based Decision Support System for Identification of COVID-19

Early-stage identification of COVID-19 authentically is a critical task for doctors. Reverse Transcription Polymerase Chain Reaction (RT-PCR) is a more reliable, helpful, and quick method of classifying and evaluating COVID- 19 cases and was widely used during the pandemic. This method does not talk about the severity level of the disease; on the other hand, medical image- based methods like Computed Tomography (CT) of the chest tell a lot about the severity level. This research paper attempts to develop a Corona Virus Identification System for Health known as COVIS-Health for COVID-19 identification using chest CT image data. In order to develop a Decision Support System, Deep learning techniques like Convolutional Neural Network (CNN) with hyper-parameters tuning, transfer learning, dual transfer learning, and augmentation were used along with Explainable Artificial Intelligence (XAI), which was added at the last layer of CNN to explain the severity level of infection. Experimental results reveal that dual transfer learning and the augmented-based VGG-19 model along with XAI outperform other CNN models by more than 99% accuracy. The developed COVIS-Health system allows end users to upload CT scan image data to identify it as COVID- 19 positive or negative. Apart from this, it also explains the severity level of any COVID-19 infected patients.