Evaluation of the Utility of Homologous Modeling and Principal Component Analysis for Sex Determination of the Mandible

Abstract

The morphology of the mandible using homologous modeling and principal component analysis, and the accuracy of sex determination based on mandibular morphology were examined. The computed tomography (CT) scans of 84 subjects (44 males, 40 females; mean age, 42.4 ± 15.4 years) were selected for this study. To avoid any effect on the morphology of the mandible, the scans of subjects with fewer than 14 remaining teeth were excluded. Homologous modeling and principal component analysis were performed using mHBM (Digital Human Techbology, Tokyo, Japan) and HBM-Rugle software (Medic Engineering, Kyoto, Japan), respectively. The contribution of the first principal component was 20.8% and that of the second principal component was 11.4%. There was a significant difference between male and female in the first principal component (Wilcoxon test, p < 0.05). Subjects with a negative first principal component value were considered more likely to be female, and those with positive values were more likely to be male (accuracy rate, 61.9%). ROC analysis of this method revealed AUC of 0.62, sensitivity of 0.48, and specificity of 0.78. Multivariate analysis was performed using all principal component values, and ROC analysis performed based on these results revealed AUC of 0.85, sensitivity of 0.82, and specificity of 0.85. Analysis using only the first principal component had lower sensitivity and specificity than reported previously, but the results using all principal component values were similar to those in past reports. This method was considered to be useful for sex determination based on mandibular morphology.