Teeth2Point: A Two-Stage Dental CBCT ROI-to-Point Segmentation Framework

Qi Ma; Shipra Jain; Niko Benjamin Huber; Ender Konukoglu

cs.CVAug 19, 2026

Teeth2Point: A Two-Stage Dental CBCT ROI-to-Point Segmentation Framework

Authors: Qi Ma, Shipra Jain, Niko Benjamin Huber, Ender Konukoglu

Organizations: Computer Vison Lab, ETH Zurich, Switzerland · Align Technology

Abstract

Modern deep learning architectures have demonstrated strong performance in dental CBCT segmentation. One remaining crucial challenge is accurate tooth labeling in cases with missing or malpositioned teeth, which are highly relevant for dental practice. Transformer-based architectures should in theory be able to resolve such ambiguities using global anatomical context. However, due to the high resolution of CBCT volumes and the wide spatial distribution of teeth within volumes, dense patch-based volumetric processing faces an inherent trade-off. Computational costs limit the number of patches that can be used in self-attention and thus, one can either increase the extent of the context captured in self-attention or capture fine-grained structural details by using small patches, but not both. In this work, we present Teeth2Point, an efficient point-based transformer framework for dental CBCT semantic segmentation that can avoid this trade-off. Teeth2Point first localizes volumetric regions of interest (ROIs) surrounding teeth using a convolutional model, then converts ROIs into point tokens using adaptive sampling. A transformer model predicts accurate segmentations using the point tokens, which allow capturing global context while retaining high resolution. The transformer is first pretrained using self-supervised learning (SSL), in the style of DINO but using domain-specific augmentation strategies, followed by supervised finetuning. The SSL pretraining, which includes random token masking, provides robustness to complex anatomical variations. Compared with the strongest two-stage baseline, Teeth2Point improves abnormal-case performance by 1.44 DSC points on average across four datasets; relative to the first-stage nnU-Net, the gain is 1.9 points.

Figures & tables

Dataset	Sample	Mean Shape	ROI Shape	FR	FR-ROI	AR
STS2024 [ 8 , 25 ]	27	(584,584,445)	(253,202,158)	0.5 $\%$	9.3%	82.5%
ToothFairy2 [ 3 , 2 , 17 ]	480	(386,348,182)	(214,169,113)	1.4%	8.3%	80.8%
ToothFairy3 [ 3 , 2 , 17 ]	532	(386,349,181)	(215,170,112)	1.3%	7.7%	23.1%
Internal Dataset	1170	(472,369,347)	(286,232,203)	0.97%	4.3%	9.0%

Table 1: Dataset Characteristics. The Region of Interest (ROI) refers to the cropped volume encompassing all foreground structures. Foreground Ratio (FR) denotes the proportion of foreground voxels in the entire volume, while FR-ROI refers to the foreground ratio within the region of interest. Abnormal Ratio (AR) measures the proportion of cases with missing or malpositioned teeth.

Figure 1: Teeth2Point Overview . The first stage uses a convolutional model for ROI extraction, followed by adaptive sampling to obtain a sparse, surface-focused representation . Point-wise predictions from the point-based transformer network are then mapped back to a volumetric form for evaluation.

Panel of Figure 2 — Figure 2: Self-distillation pretraining and Gradient-Guided Adaptive Sampling illustration. DINOv2-style pretraining is applied to large-scale unlabeled ROI points (left), while GGAS allocates more samples to high-gradient boundary regions compared to uniform sampling (right).

Method	SW	Time	STS2024		ToothFairy2		ToothFairy3		Internal		Average
Method	SW	(s)	DSC-R	DSC-A	DSC-R	DSC-A	DSC-R	DSC-A	DSC-R	DSC-A	DSC-R	DSC-A
Full-Volume CBCT
$\bullet$ nnU-Net [ 14 ]	72	12.8	97.93	93.38	97.64	93.03	97.12	93.66	94.01	87.22	96.68	91.82
Toothseg [ 24 ]	72	28.3	96.76	93.05	98.46	93.22	98.08	91.96	93.93	87.17	96.81	91.35
MedNeXt [ 22 ]	72	6.7	93.42	90.03	96.38	89.95	94.19	88.77	92.09	83.61	94.02	88.09
VideoTeeth [ 19 ]	149	51.2	96.37	94.16	97.29	91.50	94.51	91.19	94.12	85.07	95.57	90.48

Table 2: Quantitative Results . We evaluate single- and two-stage methods on STS2024, ToothFairy2/3, and an internal dataset. nnU-Net results (marked with

\bullet

) are used for second-stage ROI cropping. We report the Dice similarity coefficient (DSC) for all tooth classes and for molars separately, along with the average number of sliding windows (SW) and time for inference.

Figure 3: Visualization of pretrained representations and attention weights. We show PCA projections of feature dimensions 1–3 (a) and 4–6 (b), along with query–key attention scores in (c) and (d).

Figure 4: Qualitative Results. Teeth2Point shows strong performance among two-stage methods (numbers denote Average DSC / Molar DSC).

Figure 5: Experiments of Ablation (left) and Label Efficiency (right).

Explore similar work

Mar 12, 2026

SemiTooth: a Generalizable Semi-supervised Framework for Multi-Source Tooth Segmentation

With the rapid advancement of artificial intelligence, intelligent dentistry for clinical diagnosis and treatment has become increasingly promising. As the primary clinical dentistry task, tooth structure segmentation for Cone-Beam Computed Tomography (CBCT) has made significant progress in recent years. However, challenges arise from the obtainment difficulty of full-annotated data, and the acquisition variability of multi-source data across different institutions, which have caused low-quality utilization, voxel-level inconsistency, and domain-specific disparity in CBCT slices. Thus, the rational and efficient utilization of multi-source and unlabeled data represents a pivotal problem. In this paper, we propose SemiTooth, a generalizable semi-supervised framework for multi-source tooth segmentation. Specifically, we first compile MS3Toothset, Multi-Source Semi-Supervised Tooth DataSet for clinical dental CBCT, which contains data from three sources with different-level annotations. Then, we design a multi-teacher and multi-student framework, i.e., SemiTooth, which promotes semi-supervised learning for multi-source data. SemiTooth employs distinct student networks that learn from unlabeled data with different sources, supervised by its respective teachers. Furthermore, a Stricter Weighted-Confidence Constraint is introduced for multiple teachers to improve the multi-source accuracy.Extensive experiments are conducted on MS3Toothset to verify the feasibility and superiority of the SemiTooth framework, which achieves SOTA performance on the semi-supervised and multi-source tooth segmentation scenario.

Muyi Sun, Yifan Gao, Ziang Jia +4

School of AI, BUPT · CBDE · NLPR, CASIA +3

Medical Image Segmentation Semi-Supervised Medical Image Segmentation Cone-Beam CT

Ask PDF

Feb 25, 2026

Innovative Tooth Segmentation Using Hierarchical Features and Bidirectional Sequence Modeling

Tooth image segmentation is a cornerstone of dental digitization. However, traditional image encoders relying on fixed-resolution feature maps often lead to discontinuous segmentation and poor discrimination between target regions and background, due to insufficient modeling of environmental and global context. Moreover, transformer-based self-attention introduces substantial computational overhead because of its quadratic complexity (O(n^2)), making it inefficient for high-resolution dental images. To address these challenges, we introduce a three-stage encoder with hierarchical feature representation to capture scale-adaptive information in dental images. By jointly leveraging low-level details and high-level semantics through cross-scale feature fusion, the model effectively preserves fine structural information while maintaining strong contextual awareness. Furthermore, a bidirectional sequence modeling strategy is incorporated to enhance global spatial context understanding without incurring high computational cost. We validate our method on two dental datasets, with experimental results demonstrating its superiority over existing approaches. On the OralVision dataset, our model achieves a 1.1% improvement in mean intersection over union (mIoU).

Xinxin Zhao, Jian Jiang, Yan Tian +5

School of Computer Science and Technology, Zhejiang Gongshang University, 310018, Hangzhou, China · School of Computer and Cyber Sciences, Communication University of China, 100024, Beijing, China · Center of Big Data, China Digital Culture Group Co., Ltd, 100176, Beijing, China +6

Medical Image Segmentation

Ask PDF

Mar 17, 2026

TCATSeg: A Tooth Center-Wise Attention Network for 3D Dental Model Semantic Segmentation

Accurate semantic segmentation of 3D dental models is essential for digital dentistry applications such as orthodontics and dental implants. However, due to complex tooth arrangements and similarities in shape among adjacent teeth, existing methods struggle with accurate segmentation, because they often focus on local geometry while neglecting global contextual information. To address this, we propose TCATSeg, a novel framework that combines local geometric features with global semantic context. We introduce a set of sparse yet physically meaningful superpoints to capture global semantic relationships and enhance segmentation accuracy. Additionally, we present a new dataset of 400 dental models, including pre-orthodontic samples, to evaluate the generalization of our method. Extensive experiments demonstrate that TCATSeg outperforms state-of-the-art approaches.

Qiang He, Wentian Qu, Jiajia Dai +9

Institute of Software, Chinese Academy of Sciences · University of Chinese Academy of Sciences · Department of Computer Science and Technology, Tsinghua University +2

3D Semantic Segmentation 3D Point Cloud Segmentation 3D Medical Image Segmentation Image Segmentation

Ask PDF