Background
Traditional orthodontic treatment relies on subjective estimations of orthodontists and iterative communication with technicians to achieve desired tooth alignments, as shown in Figure 1(a). This process is time-consuming, complex, and highly dependent on the orthodontist's experience. With the development of artificial intelligence, there's a growing interest in leveraging deep learning methods to achieve tooth alignment automatically, as shown in Figure 1(b). However, the absence of publicly available datasets containing pre- and post-orthodontic 3D dental models has impeded the advancement of intelligent orthodontic solutions. To address this limitation, this paper proposes to construct the first public 3D orthodontic dental dataset. The proposed dataset is available on Zenodo1 upon request.
Methods
The construction process of orthodontic dental dataset includes four main stages: data collection, data filtering, data annotation, and data storage. The main flowchart is shown in Figure 2.
The proposed dataset was collected from 435 patients treated at Beijing Stomatological Hospital between February 2017 and June 2023. We employed the iTero intraoral scanner (Align Technology, USA) 2 to acquire patients' oral scan models (pre-orthodontic). While the digital simulation results for target tooth alignment, referred to as post-orthodontic models, are jointly designed by orthodontists and technicians. After obtaining authorization from orthodontists and patients (or with parental consent for patients under the age of 18), we collected 1,130 pairs of pre/post-treatment 3D dental models.
This paper established rigorous data inclusion criteria, requiring models to either meet Tweed’s Six Keys3 or have an ABO score4 below 30 for inclusion in the dataset. After applying these filters, we obtained a final set of 1,060 pairs of high-quality 3D dental models, including diverse malocclusion, e.g., tooth crowding, deep overbite, and deep overjet.
The data annotation process is divided into three main parts: (1) segmentation annotation of dental crowns, (2) numbering annotation of tooth position, and (3) anatomical landmarks annotation on dental crowns. The iTero intraoral scanner employed in this study automatically segments teeth and gingiva, followed by the filling of adjacent surfaces between teeth and the generation of virtual gingiva to simulate the natural oral environment. For the numbering annotation of tooth position, 2D occlusal view images were extracted from each single-arch model, and image annotation was performed using Label Studio5. Subsequently, the 2D labels were then mapped to corresponding 3D models by integrating tooth crown segmentation and tooth positions. Anatomical landmarks are critical factors in evaluating occlusion and the effect of orthodontics treatment. In the proposed dataset, we also provide anatomical landmarks annotations for dental crowns, where the landmarks for each tooth are illustrated in Figure 3. As there is currently no open-source annotation tool specifically designed for annotating the anatomical landmarks on digital dental models, we developed Fusion Analyser6, to accelerate the annotation process.
Technical Validation
To demonstrate the practical utility of the proposed orthodontic dental dataset, we provide two technical validation tasks, including tooth alignment and orthodontic effect evaluation.
Tooth Alignment Validation: In this paper, we leverage four state-of-the-art deep learning-based tooth alignment methods, i.e., TANet7, PSTN8, TAligNet9, and TADPM10, to conduct the tooth alignment experiments using our dataset. Several representative aligned dental models obtained by the above methods are demonstrated in Figure 4.
Orthodontic effect evaluation validation: In this paper, we propose several simplified occlusion evaluation metrics, based on dental crown anatomical landmarks (as shown in Figure 5), by referencing the measurement methods of the ABO-OGS index11 and DI index12. These metrics are designed to simultaneously evaluate both pre- and post-treatment dental models.
According to the results obtained by the methods mentioned above, we observed that although the alignment has improved compared to the pre-treatment models (Input), they are still far from the ground truth. This underscores the great research space for developing advanced deep-learning methods to achieve automatic tooth alignment.
References:
1. Wang, S. et al. A 3D dental model dataset with pre/post-orthodontic treatment for automatic tooth alignment [data set]. Zenodo. https://doi.org/10.5281/zenodo.11392406 (2024).
2. Wang, X. et al. Coordinate-based data analysis of the accuracy of five intraoral scanners for scanning completely dentate and partially edentulous mandibular arches. The J. Prosthet. Dent. (2024).
3. Andrews, L. F. The 6-elements orthodontic philosophy: Treatment goals, classification, and rules for treating. Am. J.Orthod. Dentofac. Orthop. 148, 883–887 (2015).
4. Song, G.-Y. et al. Validation of the american board of orthodontics objective grading system for assessing the treatment outcomes of chinese patients. Am. J. Orthod. Dentofac. Orthop. 144, 391–397 (2013).
5. Tkachenko, M. et al. Label Studio: Data labeling software (2020-2022). Open source software available from https://github.com/heartexlabs/label-studio.
6. Wang, S. et al. Fusionanalyser: A novel measurement method and software tool for dental model analysis in orthodontics. Meas. Sci. Technol. (2024).
7. Wei, G. et al. Tanet: Towards fully automatic tooth arrangement. In Eur. Conf. Comput. Vis., 481–497 (2020).
8. Li, X. et al. Malocclusion treatment planning via pointnet based spatial transformation network. In Medical Image Computing and Computer Assisted Intervention, 105–114 (2020).
9. Lingchen, Y. et al. iorthopredictor: model-guided deep prediction of teeth alignment. ACM Trans. Graph. 216 (2020).
10. Lei, C. et al. Automatic tooth arrangement with joint features of point and mesh representations via diffusion probabilistic models. Comput. Aided Geom. Des. 102293 (2024).
11. Casko, J. S. et al. Objective grading system for dental casts and panoramic radiographs. Am. J. Orthod. Dentofac. Orthop. 114, 589–599 (1998).
12. Cangialosi, T. J. et al. The abo discrepancy index: a measure of case complexity. Am. J. Orthod. Dentofac. Orthop. 125, 270–278 (2004).