Sep. 30, 1997 By Melanie Fridl Ross
GAINESVILLE, Fla.---University of Florida researchers have developed a novel computer program, a silicon soothsayer that predicts the likelihood prostate cancer has spread.
The patent-pending artificial intelligence system is notable because studies have shown surgery is not always worthwhile when cancer has progressed outside the gland or infiltrated nearby lymph nodes.
More than 300,000 men are diagnosed with prostate cancer annually in the United States; nearly half have cancer extending beyond the gland. By accurately staging prostate cancer, practitioners can help patients with potentially incurable cancer avoid radical interventions, said Dr. Ashutosh Tewari, a urology fellow at UF's College of Medicine. Treatment options include surgery, radiation therapy, cryotherapy, watchful waiting or even hormonal therapy.
Traditional tests -- such as digital rectal examination, prostate specific antigen screening, body-imaging CT or MRI scans and lymph node dissection -- are expensive and invasive and are correct only half the time, says Tewari. Tewari and Dr. Perinchery Narayan, professor and chief of urology at UF, collaborated on a pilot study to assess the computer program's accuracy.
The program, which Tewari said could save practitioners and patients $150 million in annual health-care costs, simulates the functioning of the human mind using computer processing units to assimilate information and predict outcome.
"For example, if you're crossing the street and see a car coming, you make a decision whether to cross the street or not, and if so whether you want to run or walk, based on multiple inputs," Narayan said. "They include information from your eyes, which see and assess how fast the car is coming; your muscles, which know how fast you can run or not run; and a certain portion of your brain that assesses how far you have to get before the car will or will not hit you -- what we call a past memory based on your prior experience.
"The computer uses a similar process of back-and-forth questioning that combines past memory and present data to make a determination about the cancer's severity," he said.
The computer recognizes subtle relationships among various data, then assigns a number depending on whether the patient's cancer is likely to be confined to the organ -- or 'curable' -- or whether it has spread to one or both sides of the prostate.
UF researchers spent the past four years gathering data from 1,000 prostate cancer patients at Shands at UF and the Gainesville Veterans Affairs Medical Center, including age, race and prostate specific antigen information, and entered those findings into the computer. The system then predicted whether the disease was confined to the prostate and staged it. Researchers compared the computer's results with information obtained after radical prostatectomy -- surgery to remove the prostate.
The system's so-called "negative predictive power" was very good, accurately determining cancer was confined to the prostate nine times out of 10.
UF researchers currently are working on a new network to predict the possibility of prostate cancer recurrence and life expectancy in patients who undergo treatment.
"This program will also help assess the value of newer tests," Narayan said. "By adding them to the current algorithm, we will find out whether they have additional value in terms of staging the cancer and determining prognosis, or whether traditional tests already give us the same information."
Recent UF Health Science Center news releases also are available on the UF Health Science Center Communications home page. Point your browser to http://www.vpha.health.ufl.edu/hscc/index.html
For the UF Health Science Center topic/expert list, point your browser to http://www.health.ufl.edu/hscc/experts.html
Other social bookmarking and sharing tools:
The above story is based on materials provided by University Of Florida.
Note: Materials may be edited for content and length. For further information, please contact the source cited above.
Note: If no author is given, the source is cited instead.