RESUMO
Identification of deceased with unclear identity is a common problem in forensic science, whereby radiologic comparison can be applied as method for identification. As this comparison is mostly made on a visual basis, it is highly dependent on the examiner and often lacks standardized procedures and statistical support. The aim of this study therefore was to develop a reproducible and examiner independent method for radiologic identification (RADid) based on morphometric and morphological features of sternal bone computed tomography (CT). Furthermore, the feasibility of an automated comparison of a post-mortem (PM) case against a database of ante-mortem (AM) cases was evaluated. 44 in situ PM CT scans of sternal bone and their corresponding AM CT scans were analysed and reproducible features were selected based on intra- and interrater reliability assessments. The selected features were further tested by contrasting AM and PM data. Moreover, an automated identity evaluation was developed by calculating the number of matching features between each PM case and an AM database consisting of 94 cases including the 44 with corresponding PM scans. Several features showed to be reliable according to their resulting correlation coefficient values (greater or equal to 0.60). The suitability and stability of these features was confirmed by contrasting AM and PM CT scans. Finally, the automated comparison was successful in 76.7% of the cases, whereby an unambiguous identification was possible in 65.1%. The present study reflects the benefits of a standardized and statistically established identification method and demonstrates the high potential of the sternal bone as a suitable structure for RADid.
