Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Oral Imaging for Malocclusion Issues Assessments: OMNI Dataset, Deep Learning Baselines and Benchmarking

Pujun Xue, Junyi Ge, Xiaotong Jiang, Siyang Song, Zijian Wu, Yupeng Huo, Weicheng Xie, Linlin Shen, Xiaoqin Zhou, Xiaofeng Liu, Min Gu

TL;DR

This work introduces the OMNI dataset, the first public multi-view dental image collection tailored for automated malocclusion diagnosis, comprising 4,166 RGB images from 384 participants across five views with ten annotated malocclusion categories. It benchmarks six deep learning baselines—three CNNs, two Transformers, and one GNN-based GraphTeethNet—using a COCO-format annotation pipeline and standard $mAP$ evaluation, providing a comprehensive baseline and revealing the potential of inter-tooth relational modeling. Key findings show Deformable DETR achieving strong performance, GraphTeethNet benefiting from multi-dimensional edge features, and the overall feasibility of DL methods for malocclusion localization and diagnosis on OMNI, albeit with dataset imbalance as a limitation. The dataset and benchmarks offer a valuable foundation for future dental AI research and pave the way for multi-view, automated, and scalable malocclusion assessment in clinical practice.

Abstract

Malocclusion is a major challenge in orthodontics, and its complex presentation and diverse clinical manifestations make accurate localization and diagnosis particularly important. Currently, one of the major shortcomings facing the field of dental image analysis is the lack of large-scale, accurately labeled datasets dedicated to malocclusion issues, which limits the development of automated diagnostics in the field of dentistry and leads to a lack of diagnostic accuracy and efficiency in clinical practice. Therefore, in this study, we propose the Oral and Maxillofacial Natural Images (OMNI) dataset, a novel and comprehensive dental image dataset aimed at advancing the study of analyzing dental images for issues of malocclusion. Specifically, the dataset contains 4166 multi-view images with 384 participants in data collection and annotated by professional dentists. In addition, we performed a comprehensive validation of the created OMNI dataset, including three CNN-based methods, two Transformer-based methods, and one GNN-based method, and conducted automated diagnostic experiments for malocclusion issues. The experimental results show that the OMNI dataset can facilitate the automated diagnosis research of malocclusion issues and provide a new benchmark for the research in this field. Our OMNI dataset and baseline code are publicly available at https://github.com/RoundFaceJ/OMNI.

Oral Imaging for Malocclusion Issues Assessments: OMNI Dataset, Deep Learning Baselines and Benchmarking

TL;DR

This work introduces the OMNI dataset, the first public multi-view dental image collection tailored for automated malocclusion diagnosis, comprising 4,166 RGB images from 384 participants across five views with ten annotated malocclusion categories. It benchmarks six deep learning baselines—three CNNs, two Transformers, and one GNN-based GraphTeethNet—using a COCO-format annotation pipeline and standard evaluation, providing a comprehensive baseline and revealing the potential of inter-tooth relational modeling. Key findings show Deformable DETR achieving strong performance, GraphTeethNet benefiting from multi-dimensional edge features, and the overall feasibility of DL methods for malocclusion localization and diagnosis on OMNI, albeit with dataset imbalance as a limitation. The dataset and benchmarks offer a valuable foundation for future dental AI research and pave the way for multi-view, automated, and scalable malocclusion assessment in clinical practice.

Abstract

Malocclusion is a major challenge in orthodontics, and its complex presentation and diverse clinical manifestations make accurate localization and diagnosis particularly important. Currently, one of the major shortcomings facing the field of dental image analysis is the lack of large-scale, accurately labeled datasets dedicated to malocclusion issues, which limits the development of automated diagnostics in the field of dentistry and leads to a lack of diagnostic accuracy and efficiency in clinical practice. Therefore, in this study, we propose the Oral and Maxillofacial Natural Images (OMNI) dataset, a novel and comprehensive dental image dataset aimed at advancing the study of analyzing dental images for issues of malocclusion. Specifically, the dataset contains 4166 multi-view images with 384 participants in data collection and annotated by professional dentists. In addition, we performed a comprehensive validation of the created OMNI dataset, including three CNN-based methods, two Transformer-based methods, and one GNN-based method, and conducted automated diagnostic experiments for malocclusion issues. The experimental results show that the OMNI dataset can facilitate the automated diagnosis research of malocclusion issues and provide a new benchmark for the research in this field. Our OMNI dataset and baseline code are publicly available at https://github.com/RoundFaceJ/OMNI.

Paper Structure

This paper contains 32 sections, 1 equation, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Existing dental image datasets
  4. Deep Learning for Dental Image Analysis
  5. CNN-based methods
  6. Transformer-based methods
  7. GNN-based methods
  8. OMNI Dataset
  9. Data collection protocol
  10. Participants and equipment setup
  11. Data collection scenario
  12. Relevant statistics
  13. Data privacy and ethical consideration
  14. Annotation
  15. Sample selection and data pre-processing
  16. ...and 17 more sections

Figures (5)

  • Figure 1: Multi-view malocclusion dental image acquisition process. (a) the acquisition of maxillary maxillofacial photographs; (b) the acquisition of mandibular maxillofacial photographs; (c) the acquisition of frontal occlusal photographs; and (d) the acquisition of lateral occlusal photographs.
  • Figure 2: Sample images of teeth from five different views. From left to right: frontal, right side, left side, maxillary and mandibular view.
  • Figure 3: The annotation tool and annotated examples. (a) Makesense.ai's operation page; (b) example of the annotation process; (c) the original dental image; and (d) the annotated dental image
  • Figure 4: Illustration of our GraphTeethNet baseline.
  • Figure 5: Example diagnosis results for all baseline models achieved for our OMNI dataset: (a) Faster R-CNN; (b) Mask R-CNN; (c) EfficientDet; (d) Deformable DETR; (e) DETR; (f) GraphTeethNet.