Automatic diagnosis of COVID-19 from CT images using CycleGAN and transfer learning

Navid Ghassemi, Afshin Shoeibi, Marjane Khodatars, Jónathan Heras, Alireza Rahimi, Assef Zare, Yu-Dong Zhang, Ram Bilas Pachori, Juan Manuel Górriz

Abstract: Highlights•A DL-based system is proposed for automatic detection of COVID-19 from CT-Scans.•A new dataset is collected and published from patients suspected of COVID-19.•CycleGAN is used for data augmentation alongside ViT network to reach the best results.•State-of-the-art performance, reliability, and quality of the model are shown.•Model reliability and trustworthiness are shown with calibration metrics and decision interpretation.