A Registration-Free and Slope-aware Framework for CBCT-based Dental Implant Planning

Precise planning of dental implant placement is crucial for oral rehabilitation, yet current AI-based automated methods heavily rely on accurately annotated data. Prevailing approaches obtain training labels by registering pre- and post-implantation CBCT scans, which suffers from low registration accuracy and limited paired data availability. Furthermore, implant-position prediction can benefit from explicit supervision of implant orientation to guide anatomical feature learning. To address these issues, this paper proposes RegFreeNet, a novel registration-free framework. Motivated by the clinical insight that implant positioning is largely determined by the morphology of neighboring teeth, RegFreeNet learns directly from unpaired post-implantation CBCT scans with masked implant regions, thereby eliminating the dependence on registration algorithm. However, removing implant structures introduces new challenges. The occlusion of the mandibular nerve canal within the masked region raises safety concerns regarding anatomical localization. To tackle this, we design a Neighboring Distance Perception (NDP) module, which combines multi-scale convolutions with graph-based spatial context aggregation to recover anatomical context for implant position prediction . We further introduce a slope prediction branch that shares encoder features with the implant position prediction branch. Auxiliary supervision from reference implant slopes guides shared feature learning through joint optimization of implant position and slope prediction. To validate our method, we contribute ImplantFairy, the first large-scale public dataset for dental implant planning, comprising 1,622 implant-specific CBCT samples. Extensive experiments demonstrate that RegFreeNet achieves state-of-the-art performance. Dataset access: https://github. com/sanjinYoung/ImplantFairy.

Authors

Institutions

Publication Details

Journal
IEEE Transactions on Medical Imaging
Published
2026-11-17
DOI
https://doi.org/10.1109/tmi.2026.3755031
Primary Topic
Dental Radiography and Imaging
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Registration-Free and Slope-aware Framework for CBCT-based Dental Implant Planning

Kian Ming Lim, Linlin Shen, Jianfeng Ren, Xinquan Yang et al.
IEEE Transactions on Medical Imaging
Dental Radiography and Imaging
article

A Registration-Free and Slope-aware Framework for CBCT-based Dental Implant Planning

Kian Ming Lim, Linlin Shen, Jianfeng Ren, Xinquan Yang, Xin Chen, Yongqiang Deng, Wen Ma, He Meng, Xuguang Li
article en

Abstract

Precise planning of dental implant placement is crucial for oral rehabilitation, yet current AI-based automated methods heavily rely on accurately annotated data. Prevailing approaches obtain training labels by registering pre- and post-implantation CBCT scans, which suffers from low registration accuracy and limited paired data availability. Furthermore, implant-position prediction can benefit from explicit supervision of implant orientation to guide anatomical feature learning. To address these issues, this paper proposes RegFreeNet, a novel registration-free framework. Motivated by the clinical insight that implant positioning is largely determined by the morphology of neighboring teeth, RegFreeNet learns directly from unpaired post-implantation CBCT scans with masked implant regions, thereby eliminating the dependence on registration algorithm. However, removing implant structures introduces new challenges. The occlusion of the mandibular nerve canal within the masked region raises safety concerns regarding anatomical localization. To tackle this, we design a Neighboring Distance Perception (NDP) module, which combines multi-scale convolutions with graph-based spatial context aggregation to recover anatomical context for implant position prediction . We further introduce a slope prediction branch that shares encoder features with the implant position prediction branch. Auxiliary supervision from reference implant slopes guides shared feature learning through joint optimization of implant position and slope prediction. To validate our method, we contribute ImplantFairy, the first large-scale public dataset for dental implant planning, comprising 1,622 implant-specific CBCT samples. Extensive experiments demonstrate that RegFreeNet achieves state-of-the-art performance. Dataset access: https://github. com/sanjinYoung/ImplantFairy.

IEEE Transactions on Medical ImagingVol. PP
University of Nottingham Ningbo China (CN), University of Nottingham (GB), Shenzhen University (CN), Kunming Medical University (CN)
National Natural Science Foundation of China, Ningbo Municipal Bureau of Science and Technology
Good health and well-being
Dental Radiography and Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon