Estimation of Sex from Sternum Measurements and Comparative Analysis using Machine Learning Based Feature Selection Algorithm

Abstract

This prospective study utilized multi-detector computed tomography (MDCT) to examine the morphological characteristics and sex-related variations of the sternum in 150 adult patients (92 males and 58 females) aged 20–80 years. The sternum, a skeletal component is a strong bone that doesn’t deform easily therefore, it can be used for age and sex prediction by researchers for sex determination. Sternum images were reconstructed using the Volume Rendering (VR) technique, and measurements of different sternal parts were obtained from sagittal and coronal images. Statistical analyses, including independent sample t-tests and discriminant analysis, revealed significant differences in sternum measurements between male and female groups (p<0.001), with body length and manubrium width emerging as the most crucial parameters. The study demonstrated the reliability and utility of sternal morphometric analysis for accurate sex estimation. Machine learning-based feature selection algorithms were utilized for comparative analysis of selected features. The findings emphasize the importance of employing scientifically validated methodologies, highlighting manubrium length/width and body length as key parameters for accurate gender determination using sternum measurements.

Keywords: Forensic, Manubrium, Mesosternum, Nearest neighbour classifier, ROC curve, Sternal area, Sternal cleft, Sternal foramen, Sternal index.

View PDF