Application of a novel age estimation model based on permanent maxillary canine morphometric features in a Japanese population.

OBJECTIVES: This study aims to clarify the growth curves of the long axis and apical foramen diameter of permanent maxillary canines (PMCs), and to create and evaluate a new nonlinear age estimation (NAGE) model by integrating their inverse functions within a machine learning framework. METHODS: CT measurements of length (L), apical shortest width (SW) and longest width (LW) were obtained from 726 PMCs (aged 1 to 23 years). Growth curves were modeled using Gompertz function for L and Gamma-type function for SW and LW. Inverse functions were integrated into the NAGE model: [Formula: see text]. Parameters were optimized using Soft-L1 robust least squares (SciPy, NumFOCUS, USA). Validation applied to 14 independent longitudinal cases (75 PMCs, 3-15 years). Agreement between predicted age (PA) and chronological age (CA) was evaluated using root mean squared error (RMSE), coefficient of determination (R2), and Wilcoxon signed-rank test. RESULTS: The Gompertz and Gamma-type functions effectively captured PMC growth, with RMSE of 2.5 mm (L), 1.07 mm (SW), and 1.45 mm (LW). The NAGE achieved RMSE of 1.89 years (R2 = 0.85) in the 1-23 years group, improving to 1.29 years (R2 = 0.89) in the 1-16 years group. In longitudinal cases, the overall mean difference between PA and CA was 0.12 years (1.4 months) with an RMSE of 0.99 years (R2 = 0.86) and individual errors ranging from 0.25 to 2.09 years. CONCLUSIONS: The proposed NAGE model showed promising performance for dental age estimation based on PMC length and apical foramen widths. Further studies in other teeth are needed.