Sex prediction through machine learning utilizing mandibular condyles, coronoid processes, and sigmoid notches features

Characteristics of the mandible structures have been relevant in anthropological and forensic studies for sex prediction. This study aims to evaluate the coronoid process, condyle, and sigmoid notch patterns in sex prediction through supervised machine learning algorithms. Cephalometric radiographs from 410 dental records of patients were screened to investigate the morphology of the coronoid process, condyle, and sigmoid notch and the Co-Gn distance. The following machine learning algorithms were used to build the predictive models: Decision Tree, Gradient Boosting Classifier, K-Nearest Neighbors (KNN), Logistic Regression, Multilayer Perceptron Classifier, Random Forest Classifier, and Support Vector Machine (SVM). A 5-fold cross-validation approach was adopted to validate each model. Metrics such as area under the curve (AUC), accuracy, recall, precision, and F1 Score were calculated for each model, and ROC curves were constructed. All tested variables demonstrated statistical significance (p < 0.10) and were included in the construction of the predictive model. The Co-Gn variable stood out as the most important among the evaluated independent variables, showing greater relevance in three of the four algorithms used in assessing feature importance. In the analysis of the models' performance, the AUC ranged from 0.82 [95% CI = 0.72-0.93] to 0.66 [95% CI = 0.53-0.76] for the test data, and from 0.83 [95% CI = 0.80-0.87] to 0.71 [95% CI = 0.61-0.75] for cross-validation. The precision of the models ranged from 0.83 [95% CI = 0.75-0.91] to 0.68 [95% CI = 0.58-0.78] in the test phase, and from 0.78 [95% CI = 0.74-0.82] to 0.69 [95% CI = 0.65-0.75] in cross-validation. The SVM, KNN, and Gradient Boosting Classifier algorithms stood out with the highest AUC and precision values in both cross-validation and testing. The use of condyle, coronoid process, and sigmoid notch characteristics, in combination with supervised machine learning predictive models, shows potential for contributing to sex prediction based on morphometric bone characteristics, particularly regarding the distance between the condyle and gnathion. However, given the study's limitations, these findings should be interpreted with caution.