This cross-sectional study explored the feasibility of artificial intelligence (AI)-powered tooth segmentation from cone-beam computed tomography (CBCT) for surgical splint fabrication. Digitized plaster models (0.2 mm offset) served as the reference, while AI-segmented CBCT dental arches adopted offsets of 0.25 mm, 0.5 mm, and 0.75 mm. All splints were 3D-printed and evaluated for obstruction, stability, and interfacial space (N = 15). The 0.2 mm 3D-scan splints (median 3, interquartile range (IQR) 2-3) achieved better obstruction performance than CBCT 0.25 mm counterparts (median 2, IQR 1-3, P = 0.048), with no significant differences versus CBCT 0.5 mm (median 3, IQR 3-3; P = 0.20) and 0.75 mm groups(median 3, IQR 3-3, P = 0.32). In the stability test, there was no significant difference between 3D scan 0.2 mm (median 3, IQR 2-4), CBCT 0.25 mm (median 2, IQR 1-3), CBCT 0.5 mm (median 3, IQR 3-3), and CBCT 0.75 mm (median 3, IQR 2-3). 2D and 3D interfacial space analyses revealed that 3D scan 0.2 mm splints had significantly smaller gaps (P < 0.001) than splints from AI segmentation. Collectively, AI-powered CBCT tooth segmentation with 0.5 mm and 0.75 mm offsets enables surgical splint design with acceptable performance.
Xie et al. (Wed,) studied this question.