Oguzhan Harmandaoglu, Department of Therapy And Rehabilitation, Çatalzeytin Vocational School, Kastamonu University, Kastamonu, Turkey
Yusuf Secgin, Department of Anatomy, Faculty of Medicine, Karabük University, Karabük, Turkey
Seren Kaya, Department of Anatomy, Faculty of Medicine, Duzce University, Düzce, Turkey
Deniz Senol, Department of Anatomy, Faculty of Medicine, Duzce University, Düzce, Turkey
Zulal Oner, Department of Anatomy, Faculty of Medicine, Izmir Bakırçay University, İzmir, Turkey
Omer Onbas, Department of Radiology, Faculty of Medicine, Düzce University, Düzce, Turkey
Objective: The aim of this study is to estimate gender using parameters obtained from the sphenoid sinus in computed tomography (CT) images, utilizing Machine Learning (ML) algorithms and Artificial Neural Networks (ANNs). Method: In this study, length, width, and volume measurements of the sphenoid sinus were evaluated from CT images of 300 individuals (150 males and 150 females) aged 18-65 years. Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis, Logistic Regression (LR), Extra Tree Classifier, Random Forest, Decision Tree (DT), Gaussian Naive Bayes (GaussianNB), K-Nearest Neighbors (k-NN) algorithms, and ANN model were used for gender prediction. Results: The length, width, and volume of the sphenoid sinus on both the left and right sides were found to be significantly higher in males compared to females (p < 0.05). The performance values of the ML algorithms were found as follows: LDA 0.82; k-NN 0.80; LR 0.84; GaussianNB 0.80; DT 0.82; and ANN 0.82. Conclusions: Morphometric measurements of the sphenoid sinus, when analyzed with the LDA, LR, DT, and ANN algorithms, showed high accuracy and provided reliable data for sex estimation.
Keywords: Artificial neural networks. Machine learning algorithms. Gender estimation. Sphenoid sinus.