Reproduce “MICCAI STSR 2025 Challenge: Semi-Supervised Teeth and Pulp Segmentation and CBCT-IOS Registration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yaqi Wang, Zhi Li, Chengyu Wu, Jun Liu, Yifan Zhang, Jialuo Chen, Jiaxue Ni, Qian Luo and 10 more
The MICCAI STSR 2025 Challenge established a benchmark for semi-supervised learning in dental imaging, focusing on 3D teeth and pulp segmentation in CBCT, and CBCT-IOS registration. Top-performing semi-supervised methods showed strong accuracy, with segmentation achieving an image-level Dice score of 0.9176, and dramatically reduced dental annotation time by 52–98%.
Key insights
- Semi-supervised learning (SSL) decisively validates its efficacy in overcoming data scarcity for complex dental imaging tasks.
- Hybrid deep learning architectures, including nnU-Net variants and those integrating State Space Models (e.g., Mamba), effectively capture intricate 3D anatomical details.
- High-precision AI assistance can reduce dental annotation effort by 52–98%, significantly improving clinical workflow efficiency and potentially narrowing expertise gaps among clinicians.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.