Computer-assisted dental comparison for forensic identification of human remains

Dental comparison is a technique widely used in forensic science to identify human remains. It involves the comparison of post-mortem (PM) dental information with ante-mortem (AM) information in order to establish identity. Such information includes x-ray based radiographs―including periapical film and bitewing radiographs, and panoramic radiographs or orthopantomographs (OPGs)―and written notes and charts from patient records. Dental comparison is labour-intensive, error prone, and subjective. This has motivated the research and development of computer-assisted dental comparison techniques/systems. Nevertheless the majority of existing approaches to computer-assisted dental comparison have focussed on the use of conventional x-ray radiographs because they are the most abundant source of AM image data. However, the use of PM conventional radiographs is problematic for several reasons including: their inherently poor image quality, the difficulty matching the AM and PM viewing geometries, and the difficulty obtaining such radiographs when the state of the dental remains is poor. This thesis explores an alternative approach, that does not suffer these limitations, based on acquiring PM x-ray computed tomography (CT) data and using it to reconstruct 2D radiograph-like images for matching against AM radiographs. In particular the thesis describes and develops several image processing and analysis tools for the purpose of comparing digitised AM dental radiographs to PM dental CT data. These include algorithms for synthesising/reconstructing 2D radiograph-like images from dental CT data, a metal artefact reduction algorithm for attenuating CT image artefacts resulting from metal restorations in the teeth, a semi-automatic tool for segmenting individual tooth contours in 2D radiographs, and an algorithm for automatically and quantitatively matching AM tooth contours in conventional radiographs to those in 2D panoramic-like images reconstructed from PM CT data. A prototype dental identification system (DIS), based on these algorithms, is also described and evaluated. The system is able to reconstruct a 2D panoramic-like image from a dental CT data set and to match this against a database of AM digitised bitewing radiographs to establish identity. The matching results, based on four clinical CT data sets and 32 bitewing radiographs, demonstrate the efficacy of the system and of the CT-based approach to computer-assisted dental comparison