UNIVERSITY OF MISSOURI researchers are adapting Cold War
computer technology to the war on cancer.
James Keller, a professor of electrical and computer
engineering at Mizzou, spent 12 years developing computer-vision
techniques that help the U.S. military identify tanks in the jungle.
Now he and colleagues are adapting that technology to detect
cancer in tissue biopsies and in mammograms used to screen for
"Tanks, tumors - they're both bad," Keller said in an interview
last week. "In both cases, you want to find them by separating the
object from the background."
To do that, the researchers apply a mathematical theory called
Fuzzy logic has come into its own in the past decade as a way
of solving problems where information is vague or incomplete, said
Keller, who is president of the North American Fuzzy Information
Japanese manufacturers have adapted it to make video cameras
that counter an unsteady hand and washing machines that
automatically compute how much soap and washing time any load of
Finding An Enemy Tank
Put simply, fuzzy logic is more forgiving than traditional
computer vision programs that make decisions based on strict
Take the case of an enemy tank hiding in the jungle and a
missile that seeks to make a direct hit on the tank.
A conventional computer-vision program in the incoming missile
would classify every piece of the image in the target area as part
of either a tank or a tree. But such a simple "true-false" rule
often fails to handle the complexity of the real world.
The chances for a direct hit are better with fuzzy logic,
Computer vision based on fuzzy logic allows the program to
classify pieces of the image as being "sort of like a tree" or
"sort of like a tank." It assigns a certain degree of truthfulness
to each classification, such as "I'm 30 percent confident that this
region is a tank and 80 percent confident that that region is a
The computer waits for all this evidence about the scene to be
tallied, and then makes a "judgment" about where the tank is.
This is a much more accurate way of finding the hidden target,
said Keller, who worked to develop such a computer vision system
for the Air Force and the Electronics & Space Corp. of St. Louis.
Detecting Deadly Cells
Now Keller wants to use fuzzy logic to improve ways of finding
cancerous features in the human genetic blueprint and the often
vague images of cancer tumors in the breast.
Fuzzy logic could be used to scan the human genetic blueprint
for abnormalities linked to cancer, Keller said. Such links have
begun to turn up as part of the Human Genome Project. The genome
project is the massive effort to map out all the human genes - the
genetic instructions in the blueprint for human life.
When applied to a cancer test, a fuzzy logic program would
start by scanning a biopsy of tissue from a patient where cancer is