Identifying the disappeared: testing a novel method for sorting commingled human remains

Abstract

Individualisation of commingled remains is the first step towards identification of skeletal remains, and thus to returning a loved one to their family or community after tragedy and to providing closure. Currently, there are limited methods of individualisation of commingled remains, which hinders the identification process. This dissertation explored the effectiveness of a novel individualisation technique, called mesh-to-mesh value comparison (MVC), on pair-matching four different sets of bones: humeri, clavicles, temporal bones and calcanei. This was done using two different MVC methods: one manual which uses the program Flexscan3D, and the other automatic which uses the program Viewbox. Both use three-dimensional models of bones to more accurately digitally match pairs using Iterative Closest Point algorithms. The resulting data from both programs were assessed using sensitivity, specificity, positive predictive value, and negative predictive values rates, calculated in two different manners: by Lowest Common Value selection and by Receiver Operating Characteristic (ROC) curves. Results varied by bone type and MVC type but generally both MVC methods were highly accurate at pair-matching, though the manual always outperformed the automatic. The single exception for both the manual and automatic MVC methods was the clavicle set, which produced significantly lower results.