RGU-Mamba: An U-Mamba Network with Region-Based Training Optimized for Domain Generalization Applied to Myocardial Scar and Edema Segmentation
Published in Comprehensive Analysis and Computing of Real-World Medical Images (CARE 2024), Lecture Notes in Computer Science, vol. 15548, 2025
This work presents an improved U-Mamba framework for segmenting normal myocardium, edema, and scar in cardiac MRI acquired using multiple sequences across multiple centers. Histogram matching mitigates domain shift, while a region-based training strategy captures relationships among myocardial pathologies and improves domain generalization.
Recommended citation: Gao, J., Cai, Y., Zhao, Z., Lan, X., Huang, Q., Lan, L., & Li, T.-Q. (2025). RGU-Mamba: An U-Mamba Network with Region-Based Training Optimized for Domain Generalization Applied to Myocardial Scar and Edema Segmentation. In Comprehensive Analysis and Computing of Real-World Medical Images (LNCS 15548, pp. 77–86). Springer Nature Switzerland.
Download Paper