Sex Estimation in Terms of Inclination and Alsberg in Proximal Femur by using Machine Learning Algorithms

Abstract

The fact that the femur has a solid structure ensures that the integrity of the bone is preserved, making it favorable in the sex determination process. In this study, it was aimed to predict sex by using angular variables of femur from computed tomography (CT) images and machine learning algorithms. In this study, a total of 4 angular measurements, including the femoral inclination angle (FIA) and Alsberg angle (FAA) of the proximal femur on both sides, were evaluated on CT images of 88 female and 92 male adults. Logistic regression (LR) and classification and regression tree (CART) machine learning algorithms were used for sex diagnosis. 5-fold cross-validation method was used in the training and testing processes of the models. Model performances were evaluated with area under the ROC curve, Precision and Recall statistics. Of the 4 angle measurements evaluated, only the right side FIA mean was significantly higher in women (p=0.042). The sex diagnosis success of the LR model and the CART algorithm were found to be similar (p values 0.014 and 0.017, respectively). When the success criteria of each algorithm were examined, we saw that although sex estimation was significant, (Acc 0.61, Acc 0.60, respectively) they were not very successful. We found that the machine learning algorithms applied to the variables of proximal femur angle parameters gave low accuracy of sex and the effect of both models on sex estimation was similar. © 2025 Elsevier B.V., All rights reserved.