Estimation of sex based on the sterna of Sudanese adults using multidetector computed tomography: a comparison of discriminant function analysis and binary logistic regression.

Abstract

Sex estimation is critical during forensic and anthropological investigations, and various techniques are used based on the presence of complete or fragmented human remains. This study evaluated sexual dimorphism in Sudanese sterna using multidetector computed tomography. This information was used to develop models for estimating sex, and to compare the accuracies of models based on discriminant function analysis (DFA) and binary logistic regression (BLR). The study included 126 Sudanese men and 144 Sudanese women who underwent computed tomography scans to create three-dimensional reconstructions. Six linear dimensions were measured on the manubrium and mesosternum. Men had larger mean values for most parameters, and nine parameters exhibited highly significant sexual dimorphism. The leave-one-out cross-validated sex estimation accuracies were 60.4-88.9% for DFA-based models and 60.4-89.3% for BLR-based models. The BLR-based models had noticeably better performances, with six parameters having sex estimation accuracies of >80% (vs. three parameters for DFA). The best BLR-based models incorporated the lengths and widths of the manubrium and mesosternum (accuracy: 89.3%, sex bias: 2.2%) and the combined manubrium and mesosternum lengths (accuracy: 85.6%, sex bias: 2.7%). Thus, computed tomography may be useful for measuring sternal dimensions and estimating sex among Sudanese subjects.