This is the first edition of the ToothFairy challenge organized by the University of Modena and Reggio Emilia with the collaboration of Radboud University Medical Center. The challenge is hosted by grand-challenge and is part of MICCAI2023.
Cone Beam Computed Tomography (CBCT) has become increasinglyimportant for treatment planning and diagnosis in implantdentistry and maxillofacial surgery. The three-dimensionalinformation acquired with CBCT can be crucial to plan a vast number of surgicalinterventions with the aim of preserving noble anatomical structures such as theInferior Alveolar Canal (IAC), an osseous structure of the mandible whichcontains the homonymous nerve (Inferior Alveolar Nerve, IAN), artery, and vein.Identifying the canal ensures its preservation in cases of impacted third molarextraction, implant positioning or removal of cystic lesions by preventingdamages to dental or neural structures that would significantly reduce thequality of life.
Deep learning models can support medical personnel in surgicalplanning procedures by providing a voxel-level segmentation of the IAN, whichis more accurate than bi-dimensional annotations commonly used in dailyclinical practice. Unfortunately, the small extent of available 3Dmaxillofacial datasets has strongly limited the performance of deeplearning-based techniques. On the other hand, a huge amount of sparsely2D-annotated data is produced daily in the maxillofacial practice. The incompletedetection of nerve positioning is often sufficient to facilitate a positiveoutcome of surgical intervention, but it is not an accurate anatomicalrepresentation. Nevertheless, 2D annotations fail to identify a considerableamount of inner information about the IAN position and the bone structure.Additionally, deep learning approaches frame the presence of dense 3Dannotations as a crucial factor. Still, the availability of such annotations isstrongly limited by the huge amount of time required. The challenge we proposeaims at pushing the development of deep learning frameworks to segment theinferior alveolar nerve by incrementally extending the amount of publiclyavailable 3D-annotated CBCT scans.