Comparative Evaluation of AI-Assisted and Manual CBCT-Derived Graft Volume Measurements for Maxillary Sinus Floor Augmentation

Background/Objectives: Accurate estimation of augmentation volume is essential for successful maxillary sinus augmentation planning. Manual CBCT-derived volumetric calculations remain time-consuming and operator-dependent. Artificial intelligence (AI)-assisted volumetric estimation may provide a standardized and reproducible alternative. This study evaluated the agreement and reliability of an AI-assisted volumetric estimation tool compared with CBCT-derived manually calculated augmentation volumes for sinus augmentation planning. Methods: A retrospective comparative study was conducted on 60 CBCT scans obtained from patients undergoing implant treatment planning. Radiographic measurements included bone width (A), residual alveolar bone height (B), sinus lift height (C), sinus lift width (D), and sinus lift depth (E). Manual augmentation volume was calculated using a geometric ellipsoid approximation formula derived from standardized linear measurements. AI-assisted volumetric estimation was performed using the volumetric analysis tool integrated within Planmeca Romexis 6.3 software. Two calibrated periodontists repeated both manual and AI-assisted measurements twice with a two-week interval. Reliability was assessed using intraclass correlation coefficients (ICC), while agreement between methods was evaluated using ICC and Bland–Altman analysis. Results: The mean manually calculated planned augmentation volume per implant site was 0.47 ± 0.11 cm³, whereas the mean AI-assisted planned augmentation volume was 0.52 ± 0.10 cm³. AI-assisted measurements were significantly greater than manual measurements (p < 0.001). Manual measurements demonstrated moderate intra- and inter-examiner reliability (ICC = 0.722, 0.701, and 0.649), whereas AI-assisted measurements demonstrated excellent reliability (ICC = 0.917, 0.924, and 0.903). Agreement between AI-assisted and manual volumetric estimation was good (ICC = 0.847). Bland–Altman analysis demonstrated a mean bias of 0.061 cm³ with limits of agreement ranging from 0.028 to 0.094 cm³. No significant associations were observed between augmentation volume and age, sex, or number of missing teeth after normalization per implant site (p > 0.05). Conclusions: AI-assisted volumetric estimation demonstrated excellent reproducibility and good agreement with manually calculated augmentation volumes while producing slightly higher volume estimates. AI-assisted volumetric estimation may serve as a reliable adjunctive tool for sinus augmentation planning by improving standardization and reducing operator-dependent variability.