While the paper above covers the foundation, the versioning likely refers to a specific iteration of the dataset or the nnU-Net implementation used for the challenge.
Implementation details and the submission template can be found on the AImageLab GitHub . Supplementary Reading
It utilizes 530 3D volumes (480 public, 50 private) for automated, multi-class 3D segmentation. ToothFairy 2.6.2
Focuses on improving the segmentation of the mandibular canal.
For a deeper look into the evolving methodology, you may also find these related papers relevant: While the paper above covers the foundation, the
Research on improving segmentation through advanced labeling techniques presented at CVPR.
A technical report on the specific network topology (6 resolution stages) and normalization used in the ToothFairy2 dataset. Scaling nnU-Net for CBCT Segmentation - arXiv Focuses on improving the segmentation of the mandibular
"Segmenting the Inferior Alveolar Canal in CBCTs Volumes: the ToothFairy Challenge" Journal: IEEE Transactions on Medical Imaging (2024) Key Authors: Federico Bolelli, Luca Lumetti, et al. Core Content: This paper details the first challenge (ToothFairy), including the dataset of 443 CBCT scans and a comprehensive comparative evaluation of segmentation methods for the Inferior Alveolar Canal (IAC). Key Technical Components (Version 2.6.2 Context)