Detection of Covid-19 from Chest CT Images Using Deep Transfer Learning

Abstract

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) is the virus that causes Covid-19. Covid-19 can spread quickly and lead to death so that the World Health Organization (WHO) has declared this disease a pandemic. Currently there are two methods commonly used in Covid-19, The Rapid Diagnostic Test (RDT) which has lower accuracy but requires fast time, and Real-Time Reverse Transcription Polymerase Chain Reaction (RT-PCR) which takes a long time but the accuracy is better than RDT. An alternative method that requires a short time and has high accuracy is required. One of method offered is to use CT images to detect Covid-19. This research proposes to detect Covid-19 from CT images using transfer learning methods of AlexNet, Resnet50, VGG16, Inception-v3, Inception-Resnet, Xception, and DenseNet. In this study we compared transfer learning using CLAHE preprocessing and without CLAHE. The results of this study provide that transfer learning with CLAHE preprocessing has a better performance than without CLAHE. The best performance has an accuracy of 94.97%, F-measure of 94.87%, and a precision of 97.88% for VGG16. Meanwhile, based on recall, Inception-Resnet has the best score with 95.62%, compared to VGG16 without CLAHE the results are slightly below the performance with 94.36% accuracy, F-measure of 94.21%, and a precision of 97.85, and the best recall is Resnet50 with 91.63%.