Feb. 5, 2013 at 9:46 AM ET
In some James Bond films, the spy and his enemies undergo various biometric screenings, such as retinal and handprint scans, to access important info or gain control a super secret weapon. The next time we see Bond, he could get a knee MRI to learn a covert code to stop a rogue evil weapon.
Wait, what? A knee MRI could identify someone much like a fingerprint or eye scan? It's true - Lior Shamir, a professor of computer science at Lawrence Tech University, has discovered that people’s knees are as unique as our fingerprints or eyes.
“I used to work on genetics [research] and with genetics you start thinking about what makes people different,” he says. “[This could be] something, like a fingerprint and iris, [that] is so different—our external as well as our internal [traits are unique].”
Using an algorithm, Shamir looked at a database of knee MRIs from 2,686 patients. Each person received a baseline MRI scan and a second image two years later. The algorithm looks at the pixels that make up each MRI and each scan’s unique texture and compares it to the database of images. Even though the program looks at the images at a minuscule level, Shamir notes that in some cases, even an untrained eye can match some of the knees. The program, of course, is much more reliable than the naked eye, assuring 93 percent accuracy in matching a person’s first knee image to the second.
“The accuracy cannot compete with fingerprints and iris [scans],” he says. “It’s visionary … internal body parts can be [used in] biometrics.”
Even though knees change over time, with cartilage, meniscus, and ligaments wearing down, the algorithm can still match the original picture of the knee with its newer version. While Shamir was unable to compare the knees after longer periods of time, say like what might happen over a 20-year period, he thinks that it is more difficult for people to drastically modify their knees. He believes that using a knee to definitively identify a person could be more effective in cases where people were trying to dupe the system. Anyone who has seen the movie “Seven” remembers that the serial killer sheers off all his fingertips to avoid leaving prints at the scene. While this is an extreme case, there are ways that people can trick current biometrical technology—it’s much harder to modify knees without surgery.
“The knee needs to change substantially to trick the algorithm and that is not easy to do because it involves an invasive procedure. It is not like wearing gloves or wearing sunglasses,” he explains.
While MRI identification of internal parts isn’t currently practical, if MRI technology advances quickly and becomes more affordable, it could be used to identify people at airports, for example.
The study appears in the International Journal of Biometrics.