A novel multi-attention, multi-scale 3D deep network for coronary artery segmentation

Caixia Dong, Songhua Xu, Duwei Dai, Yizhi Zhang, Chunyan Zhang, Zongfang Li

Abstract: Highlights•A novel multi-attention, multi-scale 3D deep network is proposed to comprehensively tackle the challenges of coronary artery segmentation, which is abbreviated as CAS-Net.•An attention-guided feature fusion (AGFF) module is designed to adaptively select the useful semantic information and spatial information to disentangle coronary arteries from veins and noise.•A novel scale-aware feature enhancement (SAFE) module is proposed to extract concealed multi-scale context information effectively and dynamically aggregate the multi-scale features.•A novel multi-scale feature aggregation (MSFA) module is designed to learn more semantic representations to refine the vessel segmentation map.