Computers No Better at Detecting Breast Cancer
By Serena Gordon, HealthDay Reporter
WEDNESDAY, Feb. 4 (HealthDayNews) -- Computer-aided detection systems may not be as helpful in detecting breast cancers as earlier studies had indicated.
That's the conclusion of a new study in the Feb. 4 issue of the Journal of the National Cancer Institute. It compared the results of more than 50,000 mammograms examined only by a radiologist, versus another 50,000 read by both a radiologist and a computer-aided detection system.
No statistical difference was found between the two methods. However, the study authors say more research needs to be done to determine if computer-aided detection (CAD) saves more lives and is more cost-effective than screenings using a radiologist alone.
CAD is software programmed to look for abnormalities on mammograms. In health centers where the software is in use, mammograms are screened both by the computer system and by a radiologist. CAD is not used in place of humans. In previous studies, which were done in controlled laboratory settings, it appeared that the use of CAD in addition to a radiologist's interpretation could increase the rate of cancer detection by as much as 20 percent.
"Clearly, the goal of this technology was to help detect more cancers earlier," says study co-author David Gur, a professor of radiology at Magee-Women's Hospital of the University of Pittsburgh Medical Center.
But, he says, at least in this study, the technology fell short of that goal.
"We are not saying that CAD is not good, just that in our environment, this technology did not help in detection," Gur says.
Gur and his colleagues compared three years of screening mammograms. Just over 56,400 mammograms were interpreted without the use of CAD, and 59,139 mammograms were interpreted with CAD and by a radiologist.
The mammograms were performed at the clinical breast imaging center at Magee-Women's Hospital, a high-volume academic center with highly trained radiologists on staff.
For every 1,000 mammograms studied, breast cancer detection rates were 3.6 percent without CAD and 3.5 percent with CAD. According to Gur, this difference is not statistically significant.
Dr. Joann Elmore is an associate professor of medicine and epidemiology at Harborview Medical Center at the University of Washington School of Medicine in Seattle. She finds the study results "a bit of a disappointment."
"These CAD programs have been approved by the FDA (U.S. Food and Drug Administration) and the hope is that they will assist us in catching more cancers, and this study did not show any improvement," says Elmore, who wrote an accompanying editorial in the same issue of the journal.
Both Gur and Elmore say the new study highlights the difficulty in assessing the effectiveness of new technologies. In other settings, such as a low-volume breast imaging center where the radiologists may be less experienced, the addition of CAD might be helpful in detecting more cancer, both point out.
More research needs to be done on the subject, they agree.
In the meantime, Elmore says the most important thing women need to remember is that no screening is foolproof.
"Women need to know that even with a human radiologist and a backup computer detection system, cancer can still be missed," she says.
But, she adds, it's also important for women to realize that most abnormalities discovered on mammograms turn out to be nothing. "The great majority of women who have an abnormality on a mammogram don't have breast cancer," Elmore says.
More information
For more information on mammograms, go to the National Library of Medicine. To learn more about CAD, visit the American Cancer Society.
SOURCES: David Gur, Sc.D., professor, radiology, Magee-Women's Hospital of the University of Pittsburgh Medical Center; Joann Elmore, M.D., M.P.H., associate professor, medicine and epidemiology, Harborview Medical Center at the University of Washington School of Medicine, Seattle; Feb. 4, 2004, Journal of the National Cancer Institute
Copyright � 2004 ScoutNews, LLC. All rights reserved.
|
