JEREMY PEPPAS

First came the article in the New England Journal of Medicine.
It said, "Use of CAD (Computer-Aided Detection) did not clearly improve the detection of breast cancer."

Then came the headlines in some of the country's largest newspapers. The Los Angeles Times said, "Study Finds Flaws with Computer-aided Mammograms," while on the other coast, The New York Times said, "Study Questions Exam to Detect Breast Cancer."

All those headlines for something Dr. Robert Fincher, clinical director of breast imaging for the Arkansas Cancer Research Center at UAMS, already knew.
"CAD marks so many false positives that if you based your practice of mammography on what CAD marked, you'd be biopysing everything," Fincher said. "The CAD should be the last thing that you look at."

Fincher explained how the CAD software works, "It will examine the mammogram, and it will mark anything that it finds to be unusual."

The system uses an asterisk to indicate masses and a pyramid shape to indicate a calcification.

CAD technology is less than a decade old. Fincher has been using the computer software since 1998, but he has been in business for much longer.
"I've been reading mammograms for about 30 years now," he said. "That's all I do."

In that time, Fincher estimates he has read over 150,000 mammograms.

"CAD has a lot of false positives and everyone knows that, and it does not mark every cancer either," Fincher said. He added, "I've diagnosed at least five cancers that CAD has not seen; that's why CAD will never be a reader of mammograms. It will never replace a human."

It's easy to see why.

"A human has that one thing, you can think," Fincher said. "All CAD can do is react to what it has been programmed to do. You can't program it to think."
Man 1, machine 0.

But will it always be that way?

"They are actually working on a CAD now that I think will actually be helpful," Fincher said. "It will be where the CAD can compare, like we do; it can compare the current study with a study done previously."

So in other words, the new CAD software will have a memory to recall what was marked on a mammogram five years ago and compare it to a current study. If it sees the same calcification, it would continue working.

But not anytime soon, Fincher said. "This is what they are working on, and it is not available commercially."

When it is, Fincher said, "I think (CAD) could be beneficial."

The key to the current use of CAD is experience.

"I think the people who read a lot of mammos rely less on CAD than the people who are just starting out," Fincher said. "CAD marks something and they feel obligated to mark it up, but if you've been around 30 years and reading a whole lot of mammos you'll say that CAD marked it, but it looks like it did five years ago."

Still though, Fincher said, "If CAD marks something, I'll take a quick second look at it … but the final report should not be the CAD."

Investigators at the University of California Davis Health System in Sacramento, Calif., conducted the study that the New England Journal of Medicine reported on.

In the five-year study, researchers looked at 222,135 women who had 429,345 mammograms performed in three different states.

The study found, "With the use of CAD, 32 percent more women were recalled for more tests and 20 percent more women had a breast biopsy." The authors of the study estimated "that for every additional woman diagnosed with breast cancer on the basis of CAD, 156 women are falsely recalled for more tests and 14 had unnecessary biopsies to exclude cancer."

In a much smaller local study, Fincher and his colleagues at UAMS found, "It's unusual for CAD not to mark something. We looked at 211 patients and it marked up 190 of them."

The CAD technology quickly flooded the market after its introduction in 1998, and one of the researchers on the study thought that might be part of the problem.

"There had been no large-scale community-based review of CAD efficacy despite the rapid adoption of this technology, so we did this study to see if CAD was proving to be beneficial," said lead researcher Dr. Joshua J. Fenton, MD, of the UC Davis Health System, in a press release. "It's unfortunate that the use of the software has proliferated so widely before we are certain of its benefits," said Fenton. "We need studies to determine if the benefits of the software outweigh its harms and costs. There is also the potential for new studies to improve the performance of CAD software."

But Fincher maintains, "Anyone who uses CAD to read a mammogram, that's not the way to do it." He said, "But I don't think that many people do it that way. But … there may be some people out there who don't have a lot of confidence in what they are doing and they may rely too much on CAD."

But in Little Rock and Arkansas as a whole, Fincher feels over-reliance on CAD isn't a problem.

"I know and talk to all of them," he said. "In our community, I don't think that this is an issue. It is a complementary tool. It is not a tool that we rely on to make our decision."


June 2007