Laura S. Marshall, email@example.com
CT imaging can be used not only in traditional medical applications but also to identify human remains. 3-D facial reconstruction and comparison aren’t new to forensic anthropologists, but some researchers are coming up with new programs and techniques to make the job easier.
CT scanning can be used not only to diagnose diseases but also to identify bodies. Courtesy of the Radiological Society of North America.
Dr. Leonid I. Rudin, professor of imaging science at École Normale Supérieure in Cachan, France, CEO and co-founder of Cognitech Inc. in Pasadena, Calif., and chairman of the investigative/forensic imaging conference group at SPIE, is working on using scanner-invariant slicing CT scans for facial comparison.
He began by looking at forensic anthropologists to see what they gleaned from a skull. “To us, normal people, when we look at skulls, they all look pretty much the same,” Rudin said. “They actually are not – you have to be trained.”
When you’re trained, what you look for are landmarks, or points on craniofacial surfaces where the bones come together during the development process. “If a person has high cheekbones, or deep-seated eye sockets or a protruding forehead,” Rudin said, “they can be traced back to so-called landmark points.”
His goal is to create software that can match up a 2-D photo of a person with a 3-D CT scan of the person’s skull, and it turns out that landmarks are the key. “There’s a mathematical theorem that says all you need is an alignment of three points,” he said. “They can’t be on a line – they must make a nice triangle. But three points are essential, mechanically, to determine the position of 3-D objects.
“If I have a mask or a scan, you can make a solid object based on your 3-D scan. Let’s say you have another one. All you need to do is put together the three points you think are the same. ’Cause then everything else should match, and if it doesn’t, you know you have the wrong ID.”
So to figure out which three points to match up, Rudin looked to landmarks.
“Geometry up until now doesn’t give you any way to choose those points,” he said. “Our way is very simple: On a 3-D scan, three points – if you take a triangle through them, it will be tangential to the face.
“A face typically has about 100 tri-points. And it just happened to be, with no mathematical proof of it, that the tri-points are what anthropologists call landmarks.”
Because the work is so new, Rudin tested it on subjects recruited from his wife’s doll collection, taking photographs and scanning roughly 40 dolls’ heads. He was quite pleased with the results. “Often you’ll have a doll that’s the same model but made from different molds,” he said. “We’ve been able to distinguish between molds. It’s like distinguishing between twins.”
He said the technique can be applied not only to forensics but also to security and manufacturing – for quality assurance in comparing complicated parts – and he hopes the first release for forensic scientists and anthropologists will be in the first quarter of 2009.
“Three-dimensional modeling has had, and will continue to have, a profound effect on forensic science,” said Dr. Stephanie Davy-Jow, a lecturer in forensic anthropology at Liverpool John Moores University in the UK. “Not only does it often help to create quicker and more reproducible results, but it offers the opportunity to present information visually to law enforcement, students, researchers and the public, making complicated concepts accessible and data available virtually.”
In the department of dental diagnostic science at the University of Texas at San Antonio, oral and maxillofacial radiology graduate student Joanne Ethier is applying cone-beam CT to forensic odontology – specifically, to age estimation of the deceased and to bite mark analysis.
The advantage of cone-beam CT, said Ethier, who also has a fellowship in forensic odontology and works as a consultant to the Bexar County, Texas, medical examiner, is that it’s a 3-D technique that produces no distortion or magnification, “unlike panoramic or periapical radiographs that reproduce three-dimensional structures with a two-dimension radiograph.”