A detail-oriented super-2D network for pulmonary artery segmentation

Lufei Lou, Yu Xin, Jiangbo Qian, Yihong Dong

Abstract: Highlights•We introduce DS2Net, a segmentation network based on Super-2D input.•DS2Net accurately segments with lower computational cost than 3D benchmark networks.•We design an OffsetConv algorithm for the semantic reinforcement of fine vessels.•We use the PoolFormer Module to enhance the shape-position significance of vessel features.•We use the Squeeze-and-Excitation block to build 3D correlation in the 2D network.