A Novel Region-Growing Based Semi-Automatic Segmentation Protocol for Three-Dimensional Condylar Reconstruction Using Cone Beam Computed Tomography (CBCT)

Objective: To present and validate a semi-automatic segmentation protocol to enable an accurate 3D reconstruction of the mandibular condyles using cone beam computed tomography (CBCT). Materials and Methods: Approval from the regional medical ethics review board was obtained for this study. Bilateral mandibular condyles in ten CBCT datasets of patients were segmented using the currently proposed semi-automatic segmentation protocol. This segmentation protocol combined 3D region-growing and local thresholding algorithms. The segmentation of a total of twenty condyles was performed by two observers. The Dice-coefficient and distance map calculations were used to evaluate the accuracy and reproducibility of the segmented and 3D rendered condyles. Results: The mean inter-observer Dice-coefficient was 0.98 (range [0.95–0.99]). An average 90th percentile distance of 0.32 mm was found, indicating an excellent inter-observer similarity of the segmented and 3D rendered condyles. No systematic errors were observed in the currently proposed segmentation protocol. Conclusion: The novel semi-automated segmentation protocol is an accurate and reproducible tool to segment and render condyles in 3D. The implementation of this protocol in the clinical practice allows the CBCT to be used as an imaging modality for the quantitative analysis of condylar morphology.