Segmenting the Inferior Alveolar Canal in CBCTs Volumes: The ToothFairy Challenge

Area of Science:

• Medical Imaging
• Computer Vision
• Dental Radiology

Background:

• Limited public datasets hinder comparative evaluation of Inferior Alveolar Canal (IAC) segmentation algorithms in Cone-Beam Computed Tomography (CBCT).
• This scarcity impedes deep learning research and development in dental imaging analysis.

Purpose of the Study:

• To address the lack of benchmark datasets for IAC segmentation in CBCT scans.
• To foster deep learning research and comparative evaluation through the ToothFairy challenge and a new public dataset.
• To present the details of the challenge and analyze the performance of participant algorithms.

Main Methods:

• Organized the ToothFairy challenge at MICCAI 2023, releasing a dataset of 443 CBCT scans with 153 voxel-level IAC annotations.
• Tasked participants with developing algorithms for accurate IAC identification using 2D and 3D annotated scans.
• Collected and released implementations of the best-performing algorithms in an open-source repository.

Main Results:

• The challenge facilitated the first comprehensive comparative evaluation of IAC segmentation methods on a common benchmark.
• Insights into the state-of-the-art algorithms for IAC segmentation were provided.
• The largest publicly available dataset for IAC segmentation in CBCT was established.

Conclusions:

• The ToothFairy challenge and dataset significantly advance the field of IAC segmentation in CBCT.
• The released dataset and open-source repository promote reproducibility and future research in dental AI.
• This work establishes a foundation for developing more accurate and reliable automated segmentation tools for clinical application.