[Paper Review] Automatic segmenting teeth in X-ray images: Trends, a novel data set, benchmarking and future perspectives

Gil Jader, Luciano Oliveira+1 more|arXiv (Cornell University)|Feb 9, 2018

Dental Radiography and Imaging37 references3 citations

TL;DR

This paper reviews 10 dental X-ray segmentation methods, introduces a novel public dataset of 1,500 annotated panoramic (extra-oral) X-ray images, and benchmarks performance across methods. It finds threshold-based methods dominate (54%), highlights limitations in current datasets, and advocates for deep learning to advance automated tooth segmentation in orthopantomograms.

ABSTRACT

This review presents an in-depth study of the literature on segmentation methods applied in dental imaging. Ten segmentation methods were studied and categorized according to the type of the segmentation method (region-based, threshold-based, cluster-based, boundary-based or watershed-based), type of X-ray images used (intra-oral or extra-oral) and characteristics of the dataset used to evaluate the methods in the state-of-the-art works. We found that the literature has primarily focused on threshold-based segmentation methods (54%). 80% of the reviewed papers have used intra-oral X-ray images in their experiments, demonstrating preference to perform segmentation on images of already isolated parts of the teeth, rather than using extra-oral X-rays, which show tooth structure of the mouth and bones of the face. To fill a scientific gap in the field, a novel data set based on extra-oral X-ray images are proposed here. A statistical comparison of the results found with the 10 image segmentation methods over our proposed data set comprised of 1,500 images is also carried out, providing a more comprehensive source of performance assessment. Discussion on limitations of the methods conceived over the past year as well as future perspectives on exploiting learning-based segmentation methods to improve performance are also provided.

Motivation & Objective

• To analyze and categorize state-of-the-art segmentation methods in dental X-ray imaging based on method type, image modality, and dataset characteristics.
• To address the lack of diverse, publicly available datasets for extra-oral (panoramic) X-ray images by introducing a novel, large-scale annotated dataset of 1,500 panoramic radiographs.
• To conduct a statistical benchmarking of 10 segmentation methods on the proposed dataset to evaluate performance and identify methodological limitations.
• To discuss current challenges in automated dental X-ray analysis, especially in handling image artifacts, anatomical overlaps, and low contrast.
• To explore future research directions, particularly the potential of learning-based segmentation methods such as deep learning for improved accuracy and robustness.

Proposed method

• The authors conducted a comprehensive literature review of 10 segmentation methods, classifying them by type: region-based, threshold-based, cluster-based, boundary-based, or watershed-based.
• They evaluated method performance on a newly collected dataset of 1,500 extra-oral panoramic (orthopantomographic) X-ray images, each manually annotated with tooth boundaries.
• Statistical comparison was performed using standard segmentation metrics (e.g., Dice score, Jaccard index) to assess method accuracy and robustness across the dataset.
• The study analyzed dataset variability and diversity, identifying gaps in current public datasets, particularly the over-reliance on intra-oral X-rays and limited representation of anatomical variations.
• The authors explored the potential of energy minimization and deep learning frameworks—such as fully convolutional networks (FCNs) and semantic segmentation models—for improving segmentation performance.
• They proposed that learning-based models could overcome limitations of traditional methods by automatically learning complex shape and intensity patterns from large annotated datasets.

Experimental results

Research questions

• RQ1 Which category of segmentation method is most commonly used in current dental X-ray segmentation research?
• RQ2 Do existing public datasets used to evaluate dental segmentation methods provide sufficient variability and representativeness for reliable benchmarking?
• RQ3 Which segmentation method achieves the highest performance on extra-oral panoramic X-ray images, and how do they compare statistically?
• RQ4 What are the key limitations of current segmentation techniques when applied to panoramic dental X-rays, particularly regarding image artifacts and anatomical overlap?
• RQ5 How can deep learning-based approaches advance the state of the art in automatic tooth segmentation in orthopantomograms?

Key findings

• Threshold-based segmentation methods are the most widely used, accounting for 54% of the reviewed works.
• 80% of the reviewed studies used intra-oral X-ray images, indicating a strong preference for isolated tooth regions over extra-oral panoramic views.
• The proposed dataset of 1,500 annotated panoramic X-ray images addresses a critical gap in public data availability for extra-oral dental imaging.
• Statistical benchmarking revealed significant performance variation across methods, with deep learning-based approaches showing higher potential for robustness and accuracy.
• Traditional methods struggle with image artifacts, low contrast, and overlapping anatomical structures (e.g., teeth vs. jawbone), limiting their reliability in clinical settings.
• The authors conclude that learning-based segmentation, especially deep learning, is a promising future direction for achieving more accurate and generalizable tooth segmentation in panoramic radiographs.