PDAtt-Unet: Pyramid Dual-Decoder Attention Unet for Covid-19 infection segmentation from CT-scans

Fares Bougourzi, Cosimo Distante, Fadi Dornaika, Abdelmalik Taleb-Ahmed

Abstract: Highlights•A deep architecture for segmenting Covid-19 infections in CT scans is presented.•It uses a pyramidal encoder and a dual decoder to segment infections and lungs.•The encoder is a Unet-like encoder with pyramidal paths and attention gates.•The dual decoder consists of two Att-Unet-like decoders.•Experimental results show its superiority compared to recent segmentation architectures.