Sex and stature estimation from anthropometric measurements of the foot: linear analyses and neural network approach on a Turkish sample

Abstract

For over a century, anthropometric techniques, widely used by anthropologists and adopted by medical scientists, have been utilized for predicting stature and sex. This study, conducted on a Eastern Turkish sample, aims to predict sex and stature using foot measurements through linear methods and Artificial Neural Networks. Our research was conducted on 134 medical students, comprising 69 males and 65 females. Stature and weight were measured in a standard anatomical position in the Frankfurt Horizontal Plane with a stadiometer of 0.1 cm precision. Measurements of both feet's height, length, and breadth were taken using a Vernier caliper, osteometric board, and height scale. The data were analyzed using SPSS 26.00. It was observed that all foot dimensions in males were significantly larger than in females. Sex prediction using linear methods yielded an accuracy of 94.8%, with a stature estimation error of 4.15 cm. When employing Artificial Neural Networks, sex prediction accuracy increased to 97.8%, and the error in stature estimation was reduced to 4.07 cm. Our findings indicate that Artificial Neural Networks can work more effectively with such data. Using Artificial Neural Networks, the accuracy of sex prediction for both feet exceeded 95%. Additionally, the error in stature estimation was reduced compared to the formulas obtained through linear methods.