As mammography screenings gradually have gone digital since 2000, computer-aided detection has
proliferated, too, billed as a way to find breast cancers that radiologists might have missed while
Such pattern-recognition software can, among other capabilities, pinpoint faint calcifications
in the breast that might be problems. But it has never been clear whether the technology is an
effective backstop and improves on what’s detected by the trained human eye.
An authoritative study published last month in
JAMA Internal Medicine might shift the debate. Researchers concluded that computer-aided
detection doesn’t improve diagnostic accuracy and isn’t worth the cost, which is conservatively
estimated at more than $400 million annually in the United States.
The study examined more than 625,000 mammograms interpreted by 271 radiologists between 2003 and
2009. It found that the cancer-detection rate in women screened with the pattern-recognition
software — 4.1 per 1,000 women — was no different from the rate in those who were screened without
“I’ve been really pleased at how much my colleagues in radiology have embraced this (study),”
said Dr. Constance Lehman, a study co-author and director of breast imaging in the radiology
department at Massachusetts General Hospital.
The study’s large and varied population sample of nearly 324,000 women helps address criticism
leveled at previous studies that questioned computer-aided detection. Such studies, for example,
focused on Medicare patients or were conducted when the technology was in its infancy, Lehman
“The general public is interested in how we can provide the highest-quality care at an
affordable cost,” Lehman said.
Central Ohio’s hospitals use computer-aided detection in their mammography screenings. Medicare,
for example, reimburses the Arthur G. James Cancer Hospital at Ohio State University $91.73 for a
mammogram, and an additional $5.30 for computer-assisted detection.
“This is a fairly robust study with a lot of patients,” said Dr. Jeff Hawley, interim chief of
the James’ breast radiology division.
If there’s no additional benefit from using the pattern-recognition software in screenings,
Hawley said, “should we really be using it, and is it something that should be reimbursed?”
The reimbursement issue can cut both ways, Hawley said. He said certain technologies, such as
3-D mammography, increase cancer detection, but he added that many private health-care insurance
companies don’t reimburse health-care providers that use it.
Kristen Fults-Ganey, a radiologist who reviews screening mammograms at OhioHealth Riverside
Methodist Hospital, said she and her colleagues don’t find computer-aided detection especially
However, she said that radiologists who are inexperienced in reading scans might find the backup
“I can understand if they don’t want to reimburse for it,” she said of insurers. “Nowadays, we
need to be more discriminating in what we’re paying for.”
Mount Carmel Health System will “just continue to view the evidence very closely” but does not
plan to stop using computer-assisted detection, said Mickey Johnson, the hospital system’s vice
president of women’s health.
The study didn’t evaluate whether computer-aided detection finds invasive cancers earlier in
their development, said Dr. Jeff Hoffmeister, a vice president and medical director of iCAD, a
Nashua, N.H.-based provider of advanced image-analysis software such as computer-aided
That was one conclusion of a study published in 2013 in the
Annals of Internal Medicine.
“I definitely think it is valuable,” Hoffmeister said. “When (computer-aided detection) helps
you find cancer at an earlier stage, (the cancer) costs less to treat.”
JAMA Internal Medicine study found that computer-aided detection did spot a higher rate of
ductal carcinoma in situ, a common type of noninvasive breast cancer.
“This study holds weight because of the large study, its duration, and the relatively modern
technique,” said Debra Monticciolo, chairwoman of the breast-imaging commission for the American
College of Radiology.
She said radiologists might want to review their audit data in light of the findings.
“Basing my opinion on research and my own personal use, I’m comfortable reading without