Abstract
Dental cone-beam computed tomography (CBCT) provides volumetric evidence for diagnosis and treatment planning, but construct- ing a structured report from a full volume remains challenging because clinically relevant findings are distributed across numerous slices and must be consistently linked to patient-specific tooth positions, diagnoses, and treatment actions. This challenge is amplified under limited paired supervision, where unrestricted text generation is prone to memorizing training reports or producing unsupported clinical content. To address these challenges, we propose DentSAFER, a dental self-supervised action- aware field retrieval framework with entity refinement. DentSAFER first pretrains a compact 3D encoder using variance-invariance-covariance reg- ularization and then jointly learns representations for seven report fields together with predictions for 32 FDI tooth positions and 21 treatment actions. At inference, the seven report fields are constructed through constrained retrieval from the labeled report library rather than unre- stricted decoding. During retrieval, the diagnosis and treatment plan are selected jointly using action compatibility, while patient-specific tooth codes are removed from the retrieved text and replaced with tooth codes derived from image-conditioned predictions before final report assembly. On the official test set, DentSAFER achieved a Weighted Score of 0.2193, demonstrating the feasibility of structured dental report construction under limited paired supervision.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/ODIN_challenges_006.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=N7sch3aCJR
BibTex
@InProceedings{PenRut_DentSAFER_MICCAISAT2026,
author = { Peng, Rutong AND Yang, Suhang AND Zhao, Jianxiang},
title = { { DentSAFER: Action-Aware Field Retrieval for Structured Dental CBCT Reporting } },
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}
}