Abstract
Dental surgical planning requires aligning 2D intraopera- tive intraoral radiographs (IOR) with 3D preoperative cone-beam CT (CBCT). This 2D/3D registration problem is additionally challenged by modality differences, metal artifacts, and the lack of paired IOR-CBCT data. Existing 2D/3D registration methods have focused on orthopedic anatomy with externally mounted, calibrated detectors. In contrast, IOR acquisition places the detector inside the oral cavity, requiring indepen- dent estimation of source and detector positions. We propose an end-to- end framework to estimate the 7D IOR pose relative to a known den- tal CBCT geometry. Modality-specific features are extracted using Con- vNeXt for IOR and a 3D Transformer-based model for CBCT, which are then fused via cross-attention and optimized by independent regression heads. Training uses a forward-projection pipeline generating digitally reconstructed radiographs (DRR) with known geometry from CBCTs. Compared with intensity-based and deep learning baselines, our method improved detector positioning by reducing the error from 8.0 mm to 4.4 mm, increasing the success rate from 30.9% to 71.9%, and lowering the rotation error from 2.4° to 1.4°. Similar gains were achieved for source positioning. Evaluation on real IOR-CBCT pairs showed significantly improved alignment compared to the best-performing baseline.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ODIN_003.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=BItADGmW98
BibTex
@InProceedings{RuaJos_Registration_MICCAISAT2026,
author = { Ruano-Balseca, Josué AND Joodi Bigdilo, Shojaat AND Van Leemput, Pieter AND De Beenhouwer, Jan AND Sijbers, Jan},
title = { { Registration of 2D intraoral radiographs to 3D CBCT using modality-specific encoding } },
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}
}