Algorithms Used In Military Radar Systems Have Potential to Improve Medical Ultrasonic Imaging

BUFFALO, N.Y. -- Radar systems employing image-formation algorithms developed by a University at Buffalo associate professor -- and likely being used by NATO planes to spot hidden targets in Yugoslavia -- have the potential to significantly improve medical ultrasonic imaging.

Used in the Gulf War, the imaging algorithms developed by Mehrdad Soumekh, Ph.D., associate professor of electrical engineering in the School of Engineering and Applied Sciences, produce high-resolution maps that enable users to pick out targets hidden in dense foliage or in crowded urban environments.

Soumekh recently received a $500,000 grant from the U.S. Department of Defense to establish at UB a high-performance computing center to test and refine the imaging algorithms.

But Soumekh says the system may make its greatest contributions in medical ultrasonic imaging.

"While these imaging methods have provided powerful tools for information processing in surveillance and reconnaissance radar systems, these high-resolution and high-speed systems -- and their associated algorithms -- would have a far greater impact in diagnostic medicine, leading, for example, to earlier tumor detection," he says.

Because of the high-speed nature of their measurement system, the imaging algorithms could make ultrasonic imaging up to 200 times faster than current commercial ultrasound systems without a tradeoff in resolution, an important issue in imaging dynamic targets, such as the human heart.

With funding from the National Science Foundation and defense department, Soumekh developed the algorithms, which process data measured by Synthetic Aperture Radar systems on planes flying over enemy territory. The work is described in a new book by him titled "Synthetic Aperture Radar Signal Processing" (John Wiley, 1999).

He said the military has been more willing than the medical-imaging community to accept and adapt the imaging algorithms.

Soumekh's original intent was to develop and refine what are known as wavefront reconstruction algorithms for ultrasound applications. His 1983 doctoral dissertation at the University of Minnesota focused on their application in ultrasound for the early detection of breast-cancer tumors.

At about the same time, the defense community became interested in high-resolution imaging for near-range enemy targets in reconnaissance and surveillance for military and drug-enforcement applications.

In 1992, UB patented Soumekh's algorithms and the Department of Defense began working with him in applying them for use in foliage-penetrating radar.

Since then, Soumekh's work has been funded by the U.S. Air Force and Navy, and defense department industrial contractors.

But it is in medical applications that Soumekh believes the algorithms will do the most good.

"The utility of these algorithms are particularly evident in target-imaging problems in which the conventional approximation-based methods rapidly fail, for example, in imaging near-range targets," he said. "Such problems are more often encountered in diagnostic medicine with ultrasound than in radar surveillance."

The algorithms, for example, would provide such fast generation of images that clinicians could take precise images of the human heart during surgery.

The major difference between current ultrasound systems and those envisioned by Soumekh lies in their theoretical basis. The most advanced commercial ultrasonic-imaging systems generate pictures by focusing the ultrasound beam point by point on a target region, a method that Soumekh says is prohibitively time-consuming.

By contrast, his proposed system for diagnostic ultrasonic imaging is based on a fundamental theory in optics that says that it is not necessary to create images through the conventional point-by-point method.

This theory, called the Gabor Wavefront Reconstruction Theory, developed early in the 20th century, makes it possible to exploit state-of-the-art computational power to generate fast, high-resolution images.

"With this theory, it is as though I am focusing the camera at a set of points along one line," he explained, "and all other points or targets before and after that line will appear smeared. Then I take those data, put them in the computer, and use algorithms based on my theoretical work to essentially deblur the targets."

He compares the situation to what a passenger in a car traveling quickly down a street sees when he looks at the curb.

"Your eye is making a measurement of what it sees," said Soumekh. "In ultrasound or radar, it is the same thing; the target looks blurred. But that is just the signature of the target."

According to Soumekh, his algorithms model what would be the motion of the car and the target, removing the blur from the picture and producing a signature of the target that is in focus.

"It allows us to distinguish the target from its surroundings," he said, "whether it is a man-made target hidden in foliage or a small tumor in a human organ."