Return
NucleiMix: Realistic data augmentation for nuclei instance segmentation
DOI:10.1016/j.compbiomed.2025.110901.png)
Abstract
En 中文
• Propose NucleiMix, a two-step data augmentation for realistic nuclei synthesis. • Introduce context-aware sampling to match nuclei with similar local environments. • Introduce progressive inpainting to seamlessly blend nuclei into new environments. • Evaluate on three datasets and two models, surpassing existing augmentation methods.
Keywords:
Data augmentation
Diffusion model
Nuclei instance segmentation
Pathology
Journal
IF:
6.3
Papers:
8.3K
Citations:
3.3W
Organization
No organization information available

