Self-supervised region-aware segmentation of COVID-19 CT images using 3D GAN and contrastive learning

Siyavash Shabani, Morteza Homayounfar, Varut Vardhanabhuti, Mohammad-Ali Nikouei Mahani, Mohamad Koohi-Moghadam

Abstract: Highlights•We proposed a self-supervised pipeline that does not require any manual annotation at the pixel level for segmentation.•We used GAN to synthesize healthy CT images from COVID-19 CT scans.•We used a contrastive loss to enhance the results by perform a region-aware segmentation•Our method outperforms the current state-of-the-art methods for COVID-19 CT segmentation by a reasonable amount.