Artificial Intelligence Aided Forensic Facial Identification via Craniofacial Anthropometric Measurements of Egyptian Adult Population in Sharkia Governorate

Abstract

Introduction: In forensic research studies, facial reconstruction has been presented as a human identification tool, the accuracy of which has been assessed via different subjective and objective assessment methods. These methods have been widely and empirically used in literature, but not statistically correlated. Aim of the work: The aim of the current study was to assess the role of craniofacial anthropometry as an objective method for judging the accurateness of facial reconstructions through reconstructing three dimensional faces from unknown skulls in a group of Egyptian population, then evaluating the accuracy of the results by comparing them with other subjective and objective methods. Subjects and Methods: 30 head CT scans were reconstructed via fitting facial templates onto scanned skulls. The facial reconstructions were assessed by the subjective face resemblance scores and the differences of objective surface distances. Craniofacial anthropometry is presented as an objective method for assessing the accuracy of facial reconstructions by comparing linear ratios and angles with the other methods. Results: Comparison of the ranks for each case according to the different subjective and objective tests showed no significant difference. The subjective resemblance and the objective SD tests showed the highest strongest significant correlations, and the anthropometric angles correlated with both the subjective resemblance and the objective SD tests. The anthropometric linear ratios correlated only with the subjective resemblance test. The significant correlation was stronger between the skulls-to-real faces angles differences and the skulls-to-reconstructed faces angles differences. Among the individual linear ratios, there was no specific ratio correlated with any of the other tests. The 3 angles involving points A, C and E were significantly correlated with the objective surface distance (SD) test. Conclusion: The current study provides an application of the craniofacial anthropometry as a facial identification objective method, with the craniofacial angles showed more significant correlations with the previously published subjective and objective methods. The results also showed variations among different facial parts in their predictive facial identifications values. In particular, the orbital and chin areas and their anatomical and mathematical relation to the facial width and length were significant.