Fragment distance-guided dual-stream learning for automatic pelvic fracture segmentation

Bolun Zeng, Huixiang Wang, Leo Joskowicz, Xiaojun Chen

Abstract: Highlight•A novel deep-learning method for the automatic segmentation of pelvic anatomy and fracture fragments is proposed.•A dual-stream learning strategy and a dynamic feature fusion module are developed to tackle the segmentation challenges arising from the variability of pelvic fragments.•Experiments on clinical datasets derived from two medical centers demonstrate the effectiveness and generalizability of the proposed method.•The method potentially serves as an effective tool for pelvic fracture surgical planning.