Evidence evaluation in craniofacial superimposition using likelihood ratios

Abstract

Craniofacial Superimposition is a forensic identification technique that supports decision-making when skeletal remains are involved. It is based on the analysis of the overlapping of a post-mortem skull with antemortem facial photographs. Despite its importance and wide applicability, the process remains complex and challenging. To address this, computerized methods have been proposed, but subjectivity and qualitative reporting persist in decision-making. This study introduces an evidence evaluation system proposal based on Likelihood Ratios, previously used in other forensic fields, such as DNA, voice, fingerprint, and facial comparison. We present a novel application of this framework to Craniofacial Superimposition. Our work comprises three experiments in which our LR system is trained and tested under distinct conditions concerning facial images: the first utilizes frontal facial photographs; the second employs lateral facial photographs; and the last one integrates both frontal and lateral facial photographs. In the three experiments, the proposed LR system stands out in terms of calibration and discriminating power, providing practitioners with a quantitative tool for evidence evaluation and integration. However, the lack of massive actual data obliged us to focus our study on synthetic data only. Therefore, it should be considered a proof of concept. Nevertheless, the resulting likelihood-ratio system offers objective decision support in Craniofacial Superimposition. Further studies are required to validate in a real scenario the conclusions achieved.