Machine learning-based sex estimation using palatal morphometry acquired from digital impressions.

Abstract

<h4>Background</h4>Estimating sex is a key aspect of identifying unknown human remains. Among the anatomical structures used for this purpose, the human palate is particularly valuable due to its structural integrity during the lifetime and marked sexual dimorphism, making it an interesting structure for scholars to investigate. In recent years, innovative technologies such as intraoral scanners (IOS) and machine learning (ML) have emerged as powerful tools, widely applied in the field of forensics, offering high accuracy and efficiency. This study aims to combine the previously mentioned technologies to estimate sex from morphometric features of the palate in the Iraqi subpopulation.<h4>Methods</h4>Morphometric features of the palate were extracted from digital impressions of 100 male and 100 female subjects using various software. These features were analysed to predict sex using six different ML models.<h4>Results</h4>The width of the palate in the canine area (WPC) and the area of the palate (APA) demonstrated the highest discriminative power, with areas under the curve (AUC) of 0.983 and 0.868, respectively. Among all ML models, the Support Vector Classifier (SVC) and Logistic Regression (LR) achieved the highest accuracy in sex prediction, reaching 95% as estimated by nested cross-validation (NCV).<h4>Conclusion</h4>The results indicate that the methodology used for sex estimation in this study is promising. However, further studies with larger sample sizes and inclusion of a broader range of population groups are needed to confirm and strengthen the validity of these results.