Piece by piece—a computer-aided method for virtual re-association of commingled fragmented remains

Abstract

Abstract Air crashes or explosions with numerous victims may result in thousands of fragmented human remains (FHRs) that present a massive challenge for disaster victim identification (DVI) teams. Genetic identification may present important financial and technical limits, and the physical re-association of fractured bones by forensic anthropologists may require a time-consuming phase of cleaning and drying. A virtual re-association (VRA) of FHRs using post-mortem CT-scan images could enhance the identification process and reduce the number of genetic analyses required. Therefore, this study investigated the advantages and limitations of a VRA protocol in comparison with physical re-association (PRA) in a laboratory setting and in a real case scenario. As a first step, six porcine femurs were scanned by multi-detector computed tomography (MDCT) before and after physical fragmentation. PRA of the dry bones and VRA of the 3D models of the fragments were then performed. The physically reconstructed dry bones were then once more scanned with CT. The mean distance between intact and reconstructed models, the number of re-associated fragments, and the time needed for the reconstruction were evaluated. In a second step, 87 fragmented remains resulting from a controlled pig bombing were collected, scanned, and virtually re-associated to test the feasibility of the protocol in a real context. The reconstruction of the femurs showed no difference in accuracy between PRA and VRA. Although the VRA was faster than PRA, the preparation of the material still needs to be taken into consideration. The VRA after the controlled pig bombing was limited to 8% of the total fragments. Differences in alveolar and cortical osseous structure and the presence of cartilage resulted in segmentation approximations and difficulties in the re-association itself. The explosion produced an important loss of intermediate bone elements. The VRA method still needs further evaluations with a larger sample size and different fragmentation mechanisms. However, the presented research shows promising results towards enhancing the efficiency of identifying individuals after a mass disaster.