Weakly supervised segmentation of retinal layers on OCT images with AMD using uncertainty prototype and boundary regression

Xiaoming Liu, Jia Zhang, Ying Zhang, Li Chen, Liangfu Luo, Jinshan Tang

Abstract: Highlights•A first attempt at retinal layer image segmentation using scribble supervision.•1D boundary loss and 2D reconstruction loss to mitigate topology errors.•Use uncertainty-based prototypes to alleviate the impact of interference factors.•Experimental results demonstrated the effectivity and superiority of our method.

External IDs:dblp:journals/mia/LiuZZCLT25