Sex Estimation From the Paranasal Sinus Volumes Using Semiautomatic Segmentation, Discriminant Analyses, and Machine Learning Algorithms

Abstract

Abstract The aims of this study were to determine whether paranasal sinus volumetric measurements differ according to sex, age group, and right-left side and to determine the rate of sexual dimorphism using discriminant function analysis and machine learning algorithms. The study included paranasal computed tomography images of 100 live individuals of known sex and age. The paranasal sinuses were marked using semiautomatic segmentation and their volumes and densities were measured. Sex determination using discriminant analyses and machine learning algorithms was performed. Males had higher mean volumes of all paranasal sinuses than females (P < 0.05); however, there were no statistically significant differences between age groups or sides (P > 0.05). The paranasal sinus volumes of females were more dysmorphic during sex determination. The frontal sinus volume had the highest accuracy, whereas the sphenoid sinus volume was the least dysmorphic. In this study, although there was moderate sexual dimorphism in paranasal sinus volumes, the use of machine learning methods increased the accuracy of sex estimation. We believe that sex estimation rates will be significantly higher in future studies that combine linear measurements, volumetric measurements, and machine-learning algorithms.