.png){kind=icon}
.png){kind=icon}
Return
Self-supervised small vessel segmentation with shape-aware geometric models and attention
.png){kind=icon}
PRE
AI
Z
S
J
J
J
D
L
Y
DOI：10.1016/j.media.2026.104229.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• A novel generalized explicit vessel representation. • A novel framework for self-supervised tubular structure segmentation. • Comprehensive experiments on multiple 2D and 3D datasets through image and boundary metrics for vessel segmentation assessment. • Parallel-transport-based post-processing bridges the small vessel gaps using geometrical priors. • S3U-Net outperforms other label-free vessel segmentation methods and shows its potentials on self-supervised pre-training and transfer learning tasks.
Keywords:
Small vessel representation
Shape-aware flux
Local contrast
Self-supervised learning
Journal
IF:
11.8
Papers:
3.7K
Citations:
2.4W
Organization
Cited Papers