Evaluation of the usability of facial morphometric measurements and machine learning algorithms in gender determination of adult individuals

The aim was to determine whether there is a significant difference between genders using morphometric measurements taken from anatomical structures in the Maxillary Sinus, Nasal Cavity, and Zygomatic Arch, along with machine learning algorithms. A total of 752 individuals, including 344 males and 408 females who met the inclusion and exclusion criteria, applied to the Abant İzzet Baysal University Faculty of Dentistry between 2015 and 2023, were included in the study. Morphometric measurements were performed on 12 parameters located in the maxillary sinus, nasal cavity, and zygomatic arch using the I-Cat Vision program in the Department of Oral, Dental, and Maxillofacial Radiology, Tomography Imaging, and Reporting Room. As a result of analysis with Mann-Whitney U Test; It was found that the variables of right maxillary sinus height, left maxillary sinus width and height, left nasal cavity width, left zygomatic arch width, right nasal cavity height and width, and right zygomatic arch height and width variables were statistically significantly higher in men than in women. Regardless of gender, maxillary sinus height and zygomatic arch height were higher on the right side, and nasal cavity height was higher on the left side. Random Forest, Decision Tree, K-Nearest Neighbor, Gaussian Naive Bayes, Linear Discriminant Analysis, and Support Vector Machine algorithms were used. In all three models, Linear Discriminant Analysis was the one that gave results in the shortest time. Considering the accuracy score, the Logistic Regression model was the most effective algorithm with 84% for the training set and 83% for the test set in bilateral. The best accuracy rate for measurements taken on the right side was 72% in the training data and 71% in the test data, and on the left side; The most effective algorithm in training and test data was Gaussian Naive Bayes, with 75%. It was concluded that machine learning algorithms could be used for gender determination using morphometric measurements of relevant anatomical structures.