Abstract
Supervised metal artifact reduction (MAR) for dental cone- beam computed tomography (CBCT) depends on paired metal-free and artifact-affected images, which are rarely available clinically. Existing simulation pipelines primarily model metal geometry or physical degra- dation after a metal object has been specified, leaving the upstream question of where and what restoration should be simulated underex- plored. We propose CI-MAS, a clinically informed simulation frame- work that introduces restoration configuration as an explicit decision layer before physical artifact generation. FDI-based epidemiological pri- ors are combined with patient-specific tooth semantics to determine plausible tooth sites and restoration types, which are then instanti- ated as spatially aligned metal-mask volumes and processed by cone- beam forward projection, projection-domain degradation, and FDK re- construction. Nine metal-free CBCT scans were used to generate 17 crown, filling, and mixed configurations. Under a strictly subject-disjoint train/validation/test split, retraining RISE-MAR on CI-MAS-generated data increased held-out PSNR in non-metal regions and SSIM relative to the zero-shot dental checkpoint on three unseen test subjects (PSNR 25.5 vs 23.2 dB; SSIM 0.70 vs 0.36, averaged over three training seeds). A matched tooth-agnostic random-geometry baseline achieved higher global PSNR (27.1 dB) in the current proof-of-concept setting, suggest- ing a potential trade-off between clinical configuration realism and geo- metric diversity in downstream generalization. These results support the feasibility of CI-MAS as a clinically configured CBCT artifact-simulation framework and motivate larger-scale and hybrid simulation studies.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ODIN_022.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/ODIN_022_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=TfxvDmoeI1
BibTex
@InProceedings{WanJia_CIMAS_MICCAISAT2026,
author = { Wang, Jiatao AND Li, Zhi AND Chen, Jiaqi AND Li, Zhi Han AND Sun, Jinghao AND Wen, Zhengtu AND Shang, Ping},
title = { { CI-MAS: Clinically Informed Restoration-Configuration Simulation for Dental CBCT Metal Artifact Reduction } },
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}
}
