Abstract
Accurate registration of cone-beam computed tomography (CBCT) and intraoral scans (IOS) is difficult because CBCT contains roots, bone, and metal artifacts, whereas IOS observes only crown and gingival surfaces. STSR 2026 Task 2 makes this problem semi-supervised, providing transforms for 30 cases alongside 300 unlabeled cases. We present Registration Teaches Registration (RTR). Instead of treating each labeled transform only as an output to predict, RTR uses it as a teacher: the transform places an IOS crown in CBCT and thereby re- veals where that crown is supported. These aligned surfaces form weak upper/lower targets for a semi-supervised 3D U-Net. At inference, the predicted support narrows the search over multiple IOS crown hypothe- ses; parity-aware multiscale ICP generates valid candidate transforms, and learned geometric ranking with jaw consistency selects the final pair. Learning identifies where the modalities overlap, geometry determines how they align, and ranking resolves which solution to trust. Fusing supervised and self-trained support ensembles reduces grouped target Chamfer distance from 1.118 to 1.082 mm. RTR achieves 5.7848 mm mean translation error and 2.8637◦mean rotation error on 50 validation cases and ranks first on the official hidden test.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ODIN_challenges_001.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=o1cJ9QDY6W
BibTex
@InProceedings{ZhuYi_Registration_MICCAISAT2026,
author = { Zhu, Yi AND Kechichian, Razmig AND Richert, Raphaël AND Valette, Sebastien},
title = { { Registration Teaches Registration: Transform-Derived Crown Guidance for Semi-Supervised CBCT–IOS Alignment } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17270},
month = {pending},
page = {pending}
}