Turan, Muhammed KamilOner, ZulalSecgin, YusufOner, Serkan2024-09-292024-09-2920190010-48251879-0534https://doi.org/10.1016/j.compbiomed.2019.103490https://hdl.handle.net/20.500.14619/4510Background: Predicting sex is an important problem in forensic medicine. The femur, patella, mandible and calcaneus bones are frequently used in predicting sex. In our study, we aimed to use the artificial neural network (ANN) technique to predict sex by measuring the values of the phalanges of the first and fifth toes and the first and fifth metatarsal bones. Method: All bone measurements were conducted on the direct X-ray images of 176 males and 178 females in the age range of 24-60 years. The multilayer perceptron classifier (MLPC) input layer included parameters on the bone length measurements of phalanx proximalis I, phalanx distalis I, metatarsal I, phalanx proximalis V, phalanx medialis V, phalanx distalis V and metatarsal V. The output layer contained two neurons to define the male and female sexes. The present study used an MLPC model that had two hidden layers, and the first and second hidden layers contained 14 and 7 nodes, respectively. Results: The model had an overall accuracy (Acc) of 0.95, specificity (Spe) of 0.97, sensitivity (Sen) of 0.95 and Matthews correlation coefficient (Mcc) of 0.92. While the sex prediction success of our proposed model was higher in women, the results were more specific in men and more sensitive in women (Acc(male) = 0.93, Acc(Female) = 0.98, Sen(male) = 0.93, Spe(male) = 0.98, Sen(Female) = 0.98 and Spe(Female) = 0.93). Conclusions: This study demonstrated that the ANN model for length measurements on small bones is a highly effective instrument for sex prediction.eninfo:eu-repo/semantics/closedAccessPhalanxMetatarsalX-rayArtificial neural networkMultilayer perceptron classifierSex identificationArticle10.1016/j.compbiomed.2019.1034902-s2.0-8507297025731606585Q1115WOS:000503085900004Q1