Dental records matched in just 4 minutes
LEIPZIG, Germany: It was one of the worst single aircraft accidents in the history of aviation. In August 1985, Japanese Airlines Flight 123 coming from Haneda Tokyo International Airport, suddenly went out of control and crashed at Mt. Takamagahara in the Japanese Prefecture of Gunma after 32 minutes of irregular flight. Flight data revealed a fatigue failure of the aft pressure bulkhead, followed by a structural failure of the vertical fin. The tragedy took the lives of 520 of the 524 passengers and crew members aboard the flight.
LEIPZIG, Germany: In the aftermath of the crash, over 210,000 pairs of dental radiographs had to be matched manually. Despite the help of over 2,800 forensic doctors, dentists and police officials, the identification process took more than three months. Dr Mizou Shinohara, a forensic scientist with the Gunma Prefecture Police Department at the time, helped to identify the bodies. His daughter, Dr Eiko Kosuge—who is now a Forensic Medical Examiner with the Gunma Prefecture Police Department as well and a lecturer at the Department of Oral and Maxillofacial Radiology at Kanagawa Dental College in Japan—has a strong understanding of the emotional pain the families went through while they waited to receive the remains of their loved ones due to her father’s involvement in the incident. As a result, she felt that there must be a more efficient way to identify victims through dental records.
Identification of victims in major incidents like plane crashes or natural disasters poses a tremendous amount of work for forensic scientists. With the number of fatalities on the rise, the time required for identification, as well as the risk of identification error, has increased exponentially. According to the World Disaster Report 2007 published by the International Federation of Red Cross and Red Crescent Societies, the number of disasters between 1997 and 2006 has increased by 60 per cent compared to the years from 1987 to 1996. Over the same period, the number of deaths doubled from 600,000 to 1.2 million while the average number of people affected by disasters per year rose by 17 per cent. While proof of identity is important for many reasons, such as social, legal, and financial, it is also essential for the grieving process. A positive identification and the end of hope may be devastating for the bereaved but allow for the grieving process and closure.
Among fingerprinting, medical information and DNA, dentistry is one of the most reliable methods of choice in the identification process. According to Dr Judy Hinchliffe, dental practitioner and past president of the American Society for Human Identification, “teeth are of great value for identification purposes because they can survive most conditions of nature that can damage or change other body tissues. Even in fires the posterior teeth and sometimes acrylic dentures may survive, protected by the facial and oral structures.” Good quality dental records are, therefore, essential for dental identification and for the speed with which the process can be accomplished, Dr Hinchliffe said in her article Forensic Dentistry: An Introduction to Identification Issues published in Dental Tribune Asia Pacific 11/2007.
Dr Kosuge’s idea for a new technology that would expedite the identification process began to materialize in the late 1990s. The work progressed slowly at first, but sped up with the addition of key team members such as Dr Koichi Ito, a doctor of information sciences who works as an Assistant Professor at the Graduate School of Information Sciences at Tohoku University, Japan. One of the main difficulties the project faced was that due to the nature of taking dental x-rays, distortions in the image occur. Any amount of angling of the film or x-ray tube will cause these distortions. Differences in the size and shape of the film image can cause additional problems. Early computers were capable of matching two-dimensional images such as fingerprints. However, due to the angling, rotation and distortion that occurs with the dental x-ray process, a computerized dental x-ray matching system was impossible before the development of Phase-Only Correlation (POC) technology. POC technology, which is also suitable for matching fingerprints and palm prints as well as iris scans, employs the phase information in a 2-D discrete Fourier transformation of given images. The phase information represents the object shape in images so that high-accuracy image matching using phase information can be achieved.
Kosuge’s team has developed a high-accuracy image matching technique using POC, which can estimate scaling factor, rotation angle, translational displacement and similarity between images. “We have proposed the sub-pixel correspondence technique using POC for computer vision applications,” she explains. “For dental radiograph matching we combine the above techniques to improve the accuracy of similarity evaluation.” By eliminating about 95 per cent of all possible matches the system is said to cut the time required to produce matches to about 3.6 seconds per pair and dramatically reduce the workload of forensic scientists.
The new software was already tested on single victim identifications for the Gunma Prefecture Police Department, which to date are at 100 per cent matching. In these tests, a manual identification of 60 simulated disaster victims was done alongside the new system’s procedure to check the accuracy of the results. These results were recently presented at the Annual Meeting of the Radiological Society of North America (RSNA) in Chicago, Illinois, USA. Kosuge told Dental Tribune that she and her team were very surprised and flattered by the amount of attention that the project received. “At first just being selected to be a part of the RSNA press conference was a welcome surprise, but when news of our project was picked up and carried by news agencies such as CNN and the New York Times we were a little bit shocked,” she remembered.
The team has already been contacted by various companies and individuals interested in participating in the project on various levels, from investing to collaboratively developing the software. Commercialization of the product is anticipated in about a year.