Shoulder MRI for Forensic Age Estimation: Ossification Staging of the Proximal Humeral Epiphysis

Background/Objectives: Magnetic resonance imaging (MRI) of the proximal humeral epiphysis offers a radiation-free alternative for forensic age estimation in living individuals. This study aimed to characterize the relationship between chronological age and epiphyseal maturation using a six-stage ossification classification adapted from the Schmeling-Kellinghaus framework, incorporating one original morphological criterion to distinguish the pre-terminal from the terminal stage of maturation, to establish stage-specific minimum age thresholds, and to evaluate observer reproducibility in a Turkish cohort. Methods: In this retrospective, single-center study, 395 shoulder MRI examinations (222 males, 173 females; age range, 12.17-30.67 years) obtained at Bakırköy Dr. Sadi Konuk Training and Research Hospital were evaluated. The proximal humeral epiphysis was staged on T1-weighted turbo spin echo sequences in the coronal oblique plane using a six-stage classification adapted from Schmeling et al. Two radiologists independently assessed all images; intra- and interobserver reliability were determined using Cohen's kappa. Results: Spearman's rank correlation showed a strong positive association between chronological age and ossification stage (rho = 0.825, p < 0.001; males, rho = 0.837; females, rho = 0.808). No significant sex differences in age were found at any stage (p > 0.05). The minimum age at Stage 6, the most mature stage observed, was 21.58 years in males and 21.25 years in females. Intra- and interobserver agreement were κ = 0.827 (95% CI, 0.78-0.87) and κ = 0.811 (95% CI, 0.76-0.86), respectively, corresponding to a high level of agreement. Conclusions: These findings support MRI staging of the proximal humeral epiphysis as a reliable, radiation-free adjunct for forensic age estimation, with stage-specific minimum ages providing population-specific reference values for application under the minimum age principle. Given the retrospective, single-center, single-scanner design, these results should be regarded as hypothesis-generating data pending prospective, multicenter validation.