Deformable registration network based on multi-scale features and cumulative optimization for medical image alignment

Lumin Xing, Wenjian Liu, Xing Wang, Xin Li, Rui Xu, Mini Han Wang

Abstract: Highlights•A novel unsupervised deformable registration network is developed.•A multi-scale feature fusion module is designed to extract features from different scales.•An attention mechanism optimizes the weight distribution between different spatial features.•A deformable field prediction mechanism based on cumulative learning is introduced.