Deep learning for tooth segmentation in cone-beam computed tomography: a systematic review

Introduction: Tooth segmentation in cone-beam computed tomography is a prerequisite for digital treatment planning, but manual segmentation is slow and subjective. Deep learning has been proposed to automate it. Methods: Following PRISMA 2020, PubMed, OpenAlex, and Crossref were searched for studies published from 2018 to 2025 that evaluated deep learning models for tooth segmentation in cone-beam computed tomography against a reference standard. Risk of bias was assessed with QUADAS-2. Results: Twenty-two studies were included, most using U-Net or nnU-Net architectures. The Dice similarity coefficient ranged from 0.84 for impacted canines to 0.99, with most studies reporting values above 0.90. External validation was reported in a minority of studies. Conclusions: Deep learning segments teeth in cone-beam computed tomography with high Dice coefficients in most studies, but performance varies with tooth type and dataset, and external validation remains uncommon.