Forensic personal identification utilizing part-to-part comparison of CT-derived 3D lumbar models

Abstract

The objective of this project was to document the efficacy of part-to-part comparison of computed tomography (CT)-derived three-dimensional (3D) models of the lumbar spine in forensic personal identification. By testing the methodology, this study aimed to provide a new technique of quantifiable (through a percent match) positive identification that meets the explicit requirements of the Daubert ruling and the challenges set forth in the 2009 NAS report. Ante-mortem (AM) and simulated post-mortem (PM) models of the lumbar vertebrae (L1-L5) for 30 unique individuals were compared via part comparison analyses. The threshold of ±0.5mm with at least a 90% match was considered a positive identification. Using this threshold, the part comparison results had a perfect identification rate with no false positives and no false negative matches. A ROC curve was generated with a score of 1, signifying a "perfect" sensitivity and specificity, at a cut-off value of 65.5%. On average positive IDs had a 94.7% percent match within the established threshold, while negative IDs had an average of 21.4%. In looking at the impact of different components of the biological profile, age and sex of the unknown individual played a minimal role in the percent match for both a positive and a negative ID. Lumbar level also played a minor role in in both the positive and negative percent match. The real-world application of 3D part-to-part comparison on AM and simulated PM scans demonstrate the potential usefulness of this technology in forensic identification.

Keywords: 3D; Forensics; Part-to-part comparison; Personal identification; Post-mortem computed tomography; Radiology.