Exploring sexual dimorphism in canines of contemporary North Indian populations using machine learning algorithms

Sexual dimorphism in canines of contemporary North Indian populations

Abstract

Objectives: Dentition is considered an excellent source for biological profiling in contemporary and archaeological populations with forensic anthropological, genetic, and dental perspectives. Dental dimorphism is well established and can be reflected in measurements and indices. The goal of this study is to use the discriminant function and receiver operating curve analysis to estimate sex and to make useful classification models for estimating sex based on the canine field of the mandibular and maxillary jaws.

Materials and Methods: A total of six variables of the upper and lower canines (width of left and right canines and intercanine distances) were measured on 200 adult subjects of the contemporary Haryanvi population (M/F 100:100, 18–60 years) using digital sliding calipers and indices calculated. A discriminant function and receiver operating characteristic (ROC) analysis was applied on collected data using SPSS 21.0.

Results: All variables were sexually dimorphic (p < 0.001). In stepwise analysis, maxillary intercanine distance provided an accuracy of 84%. In ROC analysis, maxillary intercanine distance emerged as an excellent variable with the maximum area under the curve (AUC) and the highest sexing accuracy (86.0%). 

Discussion: We proved the feasibility of employing machine learning to improve sex prediction. Probable causes of discrepancies in sex classification using different models are discussed. When applying models based on only canine teeth (without attachment to the tooth socket), forensic anthropologists and archaeologists should be more careful.