Abstract
Accurate segmentation of teeth and their pulp structures in cone-beam computed tomography (CBCT) is clinically important for dig- ital treatment planning, root canal therapy, and prosthetic restoration. However, manual segmentation is time-consuming and labor-intensive and depends heavily on the expertise of experienced clinicians, motivat- ing the development of fully automatic methods. We propose an nnU- Net-based global-to-local framework for tooth and pulp instance segmen- tation in the MICCAI STSR 2026 Challenge Task 1. The provided anno- tations are processed to construct solid-tooth instance targets for global localization and nested tooth–pulp region targets for local segmentation. A low-resolution nnU-Net first predicts up to 32 tooth instances from the full CBCT volume and generates tooth-wise regions of interest. A full-resolution nnU-Net then jointly segments the tooth and pulp within each region. The local predictions are subsequently mapped back to the full-volume grid, where overlapping voxels are resolved deterministically using the corresponding global localization mask. The method achieved an official validation score of 0.7643 and ranked among the top three teams on the hidden test set. These results demonstrate the effectiveness of the proposed framework for fully automatic tooth and pulp instance segmentation in dental CBCT images. The source code is publicly avail- able at https://github.com/LeguanFy/MICCAI-STSR-2026-Task1.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ODIN_challenges_022.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=BBNxVK23DU
BibTex
@InProceedings{ZhaFuy_GlobaltoLocal_MICCAISAT2026,
author = { Zhang, Fuyu AND Liu, Wei},
title = { { Global-to-Local Tooth and Pulp Instance Segmentation in Dental CBCT } },
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}
}