Computer-Aided Breast Screens Lacking
More Mammography False Positives Seen
April 4, 2007 -- Radiologists are increasingly relying on computers to help
them read screening mammograms, but the practice has not improved the detection
of invasive breast cancer, a new study shows.
Researchers concluded that computer-aided detection (CAD) is far from
enhancing breast cancer detection and has led to increases in what they termed
“potential harms,” including false positives and unnecessary biopsies.
The study was the most comprehensive analysis of computer-aided detection
ever conducted, including more than 429,000 mammograms read at 43 screening
centers located in New Hampshire, Colorado, and the state of Washington.
“We found that, among large numbers of diverse facilities and radiologists,
the use of computer software designed to improve the interpretation of
mammograms was associated with significantly higher false positive rates,
recall rates, and biopsy rates and with significantly lower overall accuracy in
screening mammography,” researcher Joshua Fenton, MD, MPH, and colleagues write
in the April 5 New England Journal of Medicine.
The promise behind computer-aided detection was that the computer programs
would help radiologists find cancers.
The computer programs read digitized versions of screening mammograms and
identify areas of concern for review by the radiologist.
One study showed that within three years of approval by the FDA in 1998,
roughly 10% of the mammography facilities in the United States were using
In an effort to better understand the benefits and risks associated with
CAD, Fenton and colleagues assessed mammography data for 222,135 women screened
between 1998 and 2002, including 2,351 women who were diagnosed with breast
cancer within one year after screening.
Seven of the 43 screening facilities included in the study began using CAD
during the study period, and outcomes from these centers after they implemented
CAD were compared with outcomes prior to its use.
The researchers concluded that CAD did not result in significant
improvements in cancer detection rates, but it did increase the number of
false-positive mammograms, resulting in significantly more patient callbacks
More Experience Needed
Robert Smith, PhD, who is director of screening for the American Cancer
Society, tells WebMD that while there are problems with CAD, “we don’t want to
throw the baby out with the bathwater.”
He points out that the radiologists at the CAD facilities in the study
tended to have less experience reading mammograms than those at the centers
where CAD was not used, and this could have influenced the findings.
“What this tells us is that the average radiologist reading mammograms needs
better training and they need more frequent feedback on what they are finding,”
he says. “Radiologists have very little opportunity to find out if they have
missed a diagnosis or have excessive callbacks that aren’t productive.”
Smith says though CAD is a potentially valuable tool, the new research makes
clear that that potential is not being fulfilled in current clinical
“A woman who is offered computer-assisted detection would be correct in
being somewhat skeptical that it would improve the likelihood of an invasive
cancer being diagnosed,” he says.