Human Identification from Computed Tomography Derived 3D Models using Part-to-Part Comparison Analysis

The identification of unknown individuals or victims is one of the most critical components of forensic death investigation. Forensic researchers are tasked with developing reproducible methods for establishing an individual's biological profile that pathologists and anthropologists can use in their casework. Radiological comparison of ante-mortem (AM) and post-mortem (PM) medical images is regularly used for identification as many individuals have undergone diagnostic scans during their lifetime. More recently, forensic practitioners have begun utilizing CT and MR imaging in conjunction with the traditional autopsy procedure. With this increased CT use in a forensic setting, the ability to do 2D and 3D comparisons via superimposition or other methods has become more widely available. One critique of radiographic matching has been the reliance on visual agreement, which can be subjective. To address the need for more quantifiable identification methods, this research attempts to validate the use of the established engineering analysis method, part-to-part comparison (or part inspection), in forensic personal identification. This study will take CT derived 3D models of the lumbar vertebrae (L 1-L5) to test their efficacy as unique identifiers. Medical image data from individuals of known demographics will be collected from the University of South Florida College of Medicine's Department of Radiology research scan database in Tampa, Florida as well as AM and PM scans from the East Midlands Forensic Pathology Unit at Leicester University. This data set will be used to examine the combined use of 3D models in conjunction with part-comparison on a large sample in concordance with reproducibility analyses. The objectives of this proposed project will be to analyze inter- and intra-observer error and reliability in 3D model creation, establish what impact the biological profile (age, sex, and ancestry) has on the percent match for identification, as well as the document the potential of part-to-part comparison as a valid forensic tool. The analysis will also be tested on PM scans to establish its accuracy in a real-world setting. By validating the methodology on actual forensic cases, this study aims 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 NAS report.

Note: This project contains a research and/or development component, as defined in applicable law, and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14).

CA/NCF