Automatic multi-anatomical skull structure segmentation of cone-beam computed tomography scans using 3D UNETR

• Published in: PloS one Vol. 17; no. 10; p. e0275033
• Main Authors: Gillot, Maxime, Baquero, Baptiste, Le, Celia, Deleat-Besson, Romain, Bianchi, Jonas, Ruellas, Antonio, Gurgel, Marcela, Yatabe, Marilia, Al Turkestani, Najla, Najarian, Kayvan, Soroushmehr, Reza, Pieper, Steve, Kikinis, Ron, Paniagua, Beatriz, Gryak, Jonathan, Ioshida, Marcos, Massaro, Camila, Gomes, Liliane, Oh, Heesoo, Evangelista, Karine, Chaves Junior, Cauby Maia, Garib, Daniela, Costa, Fábio, Benavides, Erika, Soki, Fabiana, Fillion-Robin, Jean-Christophe, Joshi, Hina, Cevidanes, Lucia, Prieto, Juan Carlos
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 12-10-2022; Public Library of Science (PLoS)
• Subjects: Artificial Intelligence; Automation; Biology and Life Sciences; Computed tomography; Computer and Information Sciences; Computer-aided medical diagnosis; Cone-Beam Computed Tomography - methods; CT imaging; Decision support systems; Deep learning; Dentistry; Head; Health aspects; Image acquisition; Image processing; Image Processing, Computer-Assisted - methods; Image segmentation; Machine learning; Medical imaging; Medicine and Health Sciences; Methods; Neural networks; Patients; Physiological aspects; Radionuclide Imaging; Research and Analysis Methods; Review boards; Skull; Skull - diagnostic imaging; Tomography; Vertebra; Vertebrae; Workflow; United States
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0275033

The segmentation of medical and dental images is a fundamental step in automated clinical decision support systems. It supports the entire clinical workflow from diagnosis, therapy planning, intervention, and follow-up. In this paper, we propose a novel tool to accurately process a full-face segmentation in about 5 minutes that would otherwise require an average of 7h of manual work by experienced clinicians. This work focuses on the integration of the state-of-the-art UNEt TRansformers (UNETR) of the Medical Open Network for Artificial Intelligence (MONAI) framework. We trained and tested our models using 618 de-identified Cone-Beam Computed Tomography (CBCT) volumetric images of the head acquired with several parameters from different centers for a generalized clinical application. Our results on a 5-fold cross-validation showed high accuracy and robustness with a Dice score up to 0.962±0.02. Our code is available on our public GitHub repository.