Integrating cervical vertebral maturation and spheno-occipital synchondrosis on CT: a non-linear framework for probabilistic age estimation.

The aim of this study was to compare the developmental timing and variability of cervical vertebral maturation (C2–C4) and spheno-occipital synchondrosis (SOS) ossification, and to evaluate these indicators using non-linear growth modelling and probabilistic age-threshold analysis to improve transparent quantification of uncertainty at legally relevant age thresholds (≥ 14, ≥ 16, and ≥ 18 years). This retrospective cross-sectional study analysed CT examinations of 986 individuals aged 10–28 years. Cervical vertebral maturation (C2–C4) and SOS ossification were assessed using five-stage ordinal systems, with age treated as a continuous variable. Skeletal maturation was examined using descriptive analyses, transition analysis, non-linear Gompertz modelling, and probabilistic age-threshold estimation. Machine learning analyses were included for comparative purposes only. Cervical vertebral maturation progressed earlier and with less variability than SOS ossification, with complete fusion most frequently observed in C2, followed by C3 and C4. SOS showed prolonged and heterogeneous maturation with substantial overlap between advanced stages. Non-linear models demonstrated earlier inflection points for cervical vertebrae (approximately 15.8–16.6 years) than for SOS (approximately 18.9 years). Advanced cervical stages were associated with high probabilities of exceeding legal age thresholds, whereas complete SOS fusion alone showed lower discriminative value, particularly for ≥ 18 years. Combined assessment yielded the highest conditional probabilities and reduced uncertainty. Machine learning results were consistent with probabilistic modelling without providing additional biologically interpretable information. Cervical vertebral maturation and SOS ossification represent distinct but complementary aspects of skeletal development. Modelling these indicators as continuous, non-linear, and probabilistic processes enables explicit management of uncertainty and improves the forensic applicability of radiological age estimation, particularly near legally relevant age thresholds.