CAD for Breast Imaging: Meeting Needs for Early Detection

 With more and more women reaching the age of regular screening mammography every year, and the fact that one in eight women will be diagnosed with breast cancer, radiologists are looking toward advanced interpretation tools to manage the volume. And, the large volumes of data generated from breast MRI means radiologists need assistance with accurate and efficient interpretation, too.

Adjunct support

“CAD should be used as an adjunct support for the radiologist after he has viewed the mammogram and reached a conclusion,” says Alan Semine, MD, medical director of The Auerbach Breast Center and chief of breast imaging at Newton-Wellesley Hospital, a member of Partners HealthCare, based in Boston, Mass. “The role of CAD is not to render judgments with regards to information but simply to alert you that you may have overlooked an observation.” In a small number of cases, CAD may not recognize an abnormality, he says, which is why it is imperative that the radiologist first completely read mammograms.

Semine says he has heard some complaints that CAD is not useful to experienced radiologists. But, “anyone can use support,” he says, “and that is exactly the role that CAD plays in the workflow.” It is important for radiologists to understand how best to use CAD. It does add time to what they’re doing but, if used properly, it should be an asset.

Semine uses ImageChecker from R2 Technology. A primary reason for implementing CAD was to eliminate the double reading process. With three radiologists performing over 800 procedures a week, CAD adds time to interpretation but double reading is extremely cumbersome, he reports. “If you can use CAD instead, it makes a lot of sense.”

Semine’s practice is growing because it is able to attract more patients outside its cachement area. That is due to “the level of service we are providing that referring physicians and patients appreciate,” he says. A challenge can be spreading the word that the advanced technology is available. “You hate to have a referring physician not be aware that you are providing a technology that a patient asks about.”

Semine also appreciates the work R2 is doing to develop new tools. “They are accomplishing some very worthwhile and constructive developments for the radiology workplace.” That includes working towards computer-aided diagnosis in the very early stages, where CAD not only alerts radiologists to certain findings but also renders a judgment regarding the level of concern that should be directed to the observation. “That shows an investment in the future that is not shying away from taking on challenges that are difficult.”

Mary M. Kelly, MD, former medical director of medical imaging at the Comprehensive Breast Center on the Swedish/Providence Campus in Seattle, and clinical associate professor at the University of Washington, uses CAD from R2 for both film and digital mammograms. The center performs about 15,000 screening mammograms a year which, on any given day, are read by one of two radiologists. Many of the screenings are done on digital equipment, either at the breast center or incorporated into a 64-foot self-contained medical coach, the "Swedish Breast Care Express." For those exams, CAD is integrated into the digital presentation. Film mammograms are also done both in the mobile setting and at the breast center; for these, films are run through a digitizer at the end of the day and CAD applied to the digitized images. When the facility implemented digital mammography at the breast center in 2003, "it made sense to give the film-screen mammography patients the same service," Kelly says. Rather quickly, CAD supplanted human being double reading for screenings there.

Kelly went with R2 because it was the company that got into CAD the earliest. Plus, she was most familiar with them. A relationship between R2 and her digital mammography equipment vendor also simplifies things, she points out.

“I find that CAD slows me down, but in a good way,” Kelly says. “The slowdown for the radiologist is trivial. It’s not an important impediment to fast throughput.” CAD essentially is the standard of care, she says. “I embrace CAD because it’s another tool to help us do a better job. We need all the help we can get in picking up breast cancer.”

A virtual consultant

Fred White, DO, radiologist and medical director of Hemet Valley Imaging in Hemet, Calif., has been using CAD from Kodak’s Health Group for about a year. The decision to invest in CAD was made for several reasons. The practice’s volume warranted the investment and “the high litigation related to mammography is something you have to consider,” he says. The practice currently is installing CR from Fuji and wanted to make the most of that investment. Plus, there is “such a high number of cancers in breasts that we don’t want to miss anything and we want to get it as early as possible.” Using CAD is like having a “virtual consultant sitting beside you” and provides an extra comfort zone with his mammography interpretations, White says.

While there is an initial learning curve, White reports, most radiologists adjust by the time they’ve read about 100 cases.

With analog mammography, technologists digitize and enter into CAD throughout the day. One of the five radiologists working at the outpatient women’s center at any given time reads the films in batches.

Since implementing Kodak CAD, Hemet has upgraded the software at least once. “Like anything in technology, it can always be improved and [the company] is continually working to improve it,” says White. “We’re happy with it. If it improves my accuracy in only one case every once in a while, it’s worth it to me.”

And White says it should be worthwhile for other facilities as well. “We conduct about 300 screening exams a month. At this volume, the payback on the CAD system is less than two years, which we consider very affordable for the added advantages it provides to our patients. I think that it’s our responsibility to do anything we can do to improve the sensitivity of our interpretation.”

Reproduceable data

David Gruen, MD, a radiologist at Norwalk Radiology & Mammography Center in Connecticut, has been using CADstream from Confirma on breast MRI for just over a year. Since breast MRI studies produce 2,000-plus images, “there is no way we could analyze in a timely manner the images of five to 10 cases per day without something to help streamline that data,” he says. “CAD helps us manage an unmanageable amount of data.”

Contrast is one of the most important aspects of breast MRI to study tumor angiogenesis. New abnormal tumor blood vessels rapidly take up and lose IV contrast. Determining the kinetics type of an abnormality is crucial. “That’s what CADstream does extraordinarily well,” says Gruen. “It takes raw data from the scanner and lets you look at the morphology of a lesion in great detail. It takes all of the enhancement data and with a few mouse clicks, lets you see whether the abnormality has type 1, 2 or 3 kinetics.” That lets Gruen give the surgeon or oncologist thorough and accurate data about what to do next.

Passing information on to other physicians makes reproduceable data vitally important, Gruen says, which was a primary reason for selecting Confirma. “We need to be able to see and process the data we see today the same way next week,” he says. “Rapid interpretation, accurate interpretation and reproduceable evaluation — you can’t ask for more than that.”

CAD standardizes evaluation and eliminates questions, Gruen says. The human eye is not that good at identifying whether something gains or loses contrast and the velocity at which it does so, he says. And that rate is very important when it comes to determining the physiology of how a tumor might be growing. “When multiple radiologists look at a mass, we would never agree,” Gruen says. “We need to objectively evaluate what the tumor is doing. CAD allowed standardization. Now, there is no question what the tumor is doing.”

The practice’s false positive rate has decreased since implementing CAD and the rate of positive biopsies has increased. “We were doing unnecessary biopsies before,” Gruen says. CAD also has offered the earliest possible detection in some cases. Gruen cites a patient with previous known breast cancer. Her mammograms were very difficult, having had multiple biopsies and extensive scarring. Her breast MRI had extensive enhancement applied for these reasons and because she was premenopausal. “On serial MRIs, we were able to detect the development of a four millimeter nodule with type 3 kinetics. We would have never picked it up without CAD and looking at different enhancement kinetics. It would have blended in with very extensive abnormal tissue.”