In this paper, we propose a novel structure-affinity dual attention-based network (SADANet) for effective spine segmentation. Global channel attention module and spatial criss-cross attention module ...