A study on sex prediction by using machine algorithms with anthropometric measurements of the seventh cervical vertebra.
Authors
Affiliations (5)
Affiliations (5)
- Department of Anatomy, Postgraduate Education Institute, Karabük University, Karabük, Türkiye.
- Department of Anatomy, Faculty of Medicine, Izmir Bakir University, Gazi Mustafa Kemal District, Kaynaklar Street, Seyrek, Menemen, Izmir, Türkiye.
- Department of Radiology, Faculty of Medicine, Izmir Bakir University, Izmir, Türkiye.
- Department of Medical Biology, Faculty of Medicine, Karabük University, Karabük, Türkiye.
- Department of Anatomy, Faculty of Medicine, Karabük University, Karabük, Türkiye.
Abstract
Prediction of sex is among important topics of forensic medicine and forensic anthropology. In studies conducted for sex prediction, pelvis and cranium bones are the most preferred bones. In cases when it is difficult to examine the pelvis and cranium bones, vertebrae have been the subject of research in sex analysis studies. The aim of this study is to predict sex by using Computed Tomography (CT) images of the vertebra prominens (C7). Another aim of the study is to make automatic measurements using labeling on C7. This retrospective study included images of 100 female and 100 male individuals (aged 2050 years). CT Images on the personal workstation (Horos Project, Version 3.0) were made orthogonal in the entire plane. They were transferred to the Sekazu program in DICOM format. The labels of the bookmarks determined on C7 were placed on the images by the Radiologist and Anatomist according to their coordinates. Then, automatic measurements were performed in the program and calculations were made. Optimization of the study was achieved by automatic measurements, thus eliminating the effects of intra-observer and/or inter-observer measurement errors. Sixteen length and 3 angle parameters were analysed by using machine learning (ML) algorithms. The accuracy rates in sex prediction using ML algorithms with the parameters obtained as a result of the analysis are as follows: Ada Boost Classification 8791%, Decision Tree 8592%, Extra Trees Classifier 8793%, Gradient Boosting Model 8591%, Gaussian Naive Bayes 8791%, Gaussian Process Classifier 8191%, K-nearest Neighbour Regression 8493%, Linear Discriminant Analysis 8894%, Linear Support Vector Classification 8892%, Non-Linear Support Vector Classification 8393%, Quadratic Discriminant Analysis 8790%, Random Forest 8392%, Support Vector Machines 8492%. In this study, it was predicted that sex prediction could be made up to 94% using ML algorithms from the parameters of vertebra prominens, which is an atypical vertebra. Therefore, we can say that vertebra prominens also shows sexual dimorphism.