CalcTalus: an online decision support system for the estimation of sex with the calcaneus and talus

The estimation of biological sex is a primary source of information regarding unidentified skeletal individuals in bioarcheological and forensic contexts. This study aims to propose new metric standards for the estimation of sex using variables of the calcaneus and talus. An ancillary goal encompasses the creation of a web-based decision support system for the assessment of sex. Six measurements from the talus and nine from the calcaneus were collected from 180 adult individuals (93 females; 87 males) belonging to the Coimbra Identified Skeletal Collection. Logistic regression (LR), support vector machines (SVM), and a decision-tree algorithm were employed to develop models for sex prediction. Univariable sectioning points generated with a decision-tree algorithm yielded an accuracy under cross-validation from 78.3 to 82.2% with talar measurements, and from 73.6 to 86.4% with calcanei variables. Systematic error ranged from 0.2 to 34.1%. Univariable and multivariable models, produced with LR and SVM, correctly predicted sex in 85.0-91.3% of cases (bias from 0.3 to 4.3%). Obtained cross-validated accuracies obtained with the new models are similar to earlier results on the subject. The performance of multivariable model predictive is substantially superior, hinting the relevance of population-specific standards for sex estimation. The operationalization of these models in a free, user-friendly, web-application-CalcTalus (http://osteomics.com/CalcTalus/)-facilitates the probabilistic assessment of sex, providing performance metrics for the statistical templates.