Medical image segmentation based on dynamic positioning and region-aware attention

Zhongmiao Huang, Shuli Cheng, Liejun Wang

Abstract: Highlights•A two-stage network is proposed for the differences in deep and shallow semantic information.•Designs dynamic positioning attention (DPA) to enhance the learning of local information.•Uses Bi-Level Routing Attention (BRA) to improve computational efficiency.•Numerous experiments on multiple datasets have demonstrated the effectiveness of our method.