Age and sex estimation using intracranial radiographic biometric measurements

Identification is the process of determining the unique characteristics of individuals, with age and sex estimation constituting its fundamental components. This study aimed to determine age and sex by retrospectively analyzing axial and sagittal computed tomography images (1 mm slice thickness) of 822 individuals (431 males, 391 females) aged 18 to 25 years. In this age range, considered a 'grey area' in the literature, a total of 32 morphometric parameters were evaluated, including 13 direct measurements, 17 calculated ratios, and two asymmetry indices derived from the cranial region, atlas, and dens axis structures. For sex classification, partial least squares discriminant analysis and logistic regression models were established using 24 statistically significant parameters. To reduce model complexity, simplified partial least squares discriminant analysis and discriminant function analysis models were developed with the three most valuable parameters based on variable importance in projection scores. For continuous age estimation, a main partial least squares regression model with 17 parameters and a simplified model including the top three correlating variables were constructed; however, the predictive performance of both models was found to be insufficient. Moreover, no statistically significant differences were observed between all age subgroups (18-19, 20-21, 22-23, and 24-25 years), precluding the development of an age classification model. In conclusion, a high-performance, three-parameter, and directly applicable partial least squares discriminant analysis model for sex estimation was developed (overall accuracy: 82,5 85,4%). It is recommended that this model's generalizability be further evaluated in larger, multi-center studies involving diverse ethnic populations.