RESEARCH Science Engineering

Fast algorithm could aid luggage inspection and medical imaging

James E. Kloeppel, Physical Sciences Editor
(217) 244-1073; Kloeppel@uiuc.edu

2/12/03

CHAMPAIGN, Ill. — Beleaguered luggage scanners at the nation’s airports may soon find help in a fast algorithm developed by scientists at the University of Illinois at Urbana-Champaign. The algorithm also promises to speed delivery of images generated by computerized tomography in hospitals and industry.

Computerized tomography is commonly used to create cross-sectional images from many individual slices – or scans – illuminated by X-rays. To produce an image, the data from a CT scanner usually needs to be processed in a complicated and time-consuming manner. The new algorithm uses fast hierarchical backprojection and reprojection methods to greatly improve the speed at which these images are reconstructed.
"The computational savings for a typical medical image is roughly fiftyfold with our algorithm," said Yoram Bresler, a professor of electrical and computer engineering at Illinois. "As the image size goes up, the computation for any algorithm also goes up, but our computational savings increases with image size."

Faster imaging speeds could offer dramatic improvements in the three-dimensional, X-ray inspection of checked luggage at the nation’s busiest airports.

"Current X-ray scanners are not actually performing three-dimensional volumetric imaging," said David Munson Jr., the Robert C. MacClinchie Distinguished Professor of Electrical and Computer Engineering at Illinois. "Instead, these systems are reconstructing only a small number of slices through the volume, which makes it harder to detect weapons, explosives, or other hazardous materials."

Three-dimensional, CT image analysis would be far more effective, Munson said. "As baggage moves along a conveyor belt at high speed and high volume, the computer would quickly collect the data, produce and analyze the images. Suspicious pieces of luggage could be sent to a human operator for a more thorough analysis."

In the medical field, computerized tomography is widely used for CT scans and for positron emission tomography. With positron emission tomography – which is used to study the functional operation of the brain – the image is built gradually through a reiterative process. The fast algorithm could speed this process dramatically.

"In some medical applications, a physician will modify his actions based upon the images he is obtaining," Bresler said. "While the data in CT diagnostic imaging can be acquired quickly, reconstructing the image with conventional algorithms can create a significant time delay."

To watch as a catheter is carefully threaded through a beating heart, or to perform image-guided surgery of the brain, for example, any time lag in image reconstruction is unacceptable, Bresler said. "Our fast reconstruction algorithm would enable doctors to watch movements and perform corrective procedures in real time."

The fast algorithm also could assist industry, where computerized tomography is used for nondestructive testing and evaluation. Faster image processing could help improve the detection of defects in jet-engine turbine blades, for example. In the lumber industry, three-dimensional CT scanners could evaluate logs for internal defects and determine how to cut the logs for maximum yield.

"The new algorithm works with both the two-dimensional, fan-beam geometry used in existing commercial CT scanners, and with the three-dimensional, cone-beam geometry that will be used in next-generation machines," Munson said. "Because the algorithm works without the expensive, special-purpose hardware required with other backprojection techniques, substantial capital costs can be saved."

Collaborators include electrical and computer engineering professor Eric Michielssen, visiting scientist Amir Boag, and graduate students Shu Xiao and Samit Basu. The researchers have applied for a patent.