Sex estimation from first rib 3D morphometry in a Thai population: Establishing baseline standards through comparative evaluation of traditional statistical and machine learning approaches.

Sex estimation from skeletal remains is fundamental to forensic identification. The first rib demonstrates enhanced preservation through superior cortical density and protected positioning, yet population-specific standards for Thai populations remain absent. To establish the first sex estimation standards from first rib morphometry in Northeastern Thai populations and evaluate comparative performance of discriminant function analysis (DFA), binary logistic regression (BLR), support vector machine (SVM), and Random Forest (RF). Six hundred first ribs from 300 males and 300 females with mean age of 56.67 ± 13.88 years from documented skeletal collections were digitized using 3D scanning. Seven morphometric parameters were measured by two trained observers with inter-observer reliability assessment. Four classification models were developed and their performance compared through cross-validation and ROC analysis. Inter-observer reliability was excellent with intraclass correlation coefficients ranging from 0.992 to 0.999. All parameters demonstrated significant sexual dimorphism at p < 0.01 with Cohens d ranging from 0.57 to 1.38. Length from tubercle to rib head with d = 1.38, distance from tubercle to internal margin with d = 1.17, and weight with d = 0.94 showed strongest discrimination. Comparative analysis revealed overall accuracy ranging from 75.2% to 81.0%. Random Forest achieved highest performance with 81.0% accuracy and AUC of 0.867 alongside 4.6% sex disparity. BLR demonstrated 80.3% accuracy with perfect balance showing 0% disparity, while DFA yielded 80.2% with 2.3% disparity. SVM underperformed at 75.2% with substantial bias of 19.0% disparity. This study provides the first validated sex estimation standards for Thai first rib morphometry. Binary logistic regression emerges as optimal for forensic application, balancing 80.3% accuracy, perfect sex-specific performance, and methodological transparency essential for legal proceedings, despite Random Forests marginally superior discriminatory capacity.