Digital Mammography CAD: A Second Pair of Eyes that Makes a Difference

With the sharp surge in digital mammography system implementations, mammography CAD is riding the curve, too. This “second set of eyes” that helps to confirm radiological findings in mammograms, makes sense from both a cost perspective and the potential it offers to detect more breast lesions, according to many radiologists.

The newer systems coming on line—including those from Hologic and iCAD—are satisfying radiologists’ needs for fewer false positives, and relieving fears that were roused in April by a New England Journal of Medicine article alleging that CAD had, at best, a neutral impact on detection of lesions, and at worst, was missing lesions that otherwise might be spotted. 

Proponents of digital mammography CAD believe the technology increases the detection of early breast cancers, especially those in women with dense breast tissue and that the technology has evolved since the data examined in the study. James Ruiz, MD, a radiologist with Woman’s Hospital in Baton Rouge, La., believes that the New England Journal of Medicine study was poorly designed, implying that radiologists would substitute the second opinion of a colleague for CAD, potentially missing diagnosis that would otherwise be made.

A user of Hologic’s R2 ImageChecker, Ruiz and his colleagues use a triple-review process, whereby two radiologists automatically read every mammogram and also view the CAD findings. “We have found that detection of a lesion by a second reader has decreased since we started using CAD as the third reader because the CAD system is picking up some of those lesions that the second radiologist used to pick up,” he says. “As applied to our practice, CAD is helping us pick up cancers that we might otherwise have missed.” Woman’s Hospital has six radiologists on staff.

Alan Melton, MD, an assistant clinical professor of radiology at Columbia University in New York, who specializes in mammography, believes that CAD is a useful assist for radiologists in general practice who don’t specialize in mammography; the technology also is valuable as a back-stop for radiologists specializing in mammography by providing extra clarity especially as the technology has evolved. Melton uses GE Senographe Workstations with iCAD SecondLook Digital CAD.


Newer algorithms boost accuracy

Because of the differing algorithms that make up the various CAD systems on the market, one of the biggest problems facing designers of those algorithms and radiologists has been the high number of false positive marks on the CAD interpretation. The newer systems that are just getting approved and coming online at radiologists’ offices across the country are making significant strides in reducing the number of false positives marked by the technology.

Even with systems using prior-generation algorithms, digital CAD technology still prompts radiologists to take a second look or bring a patient back for further tests, holding out some promise of detecting some findings that a radiologist may miss. Other potential benefits of digital CAD include improved reader sensitivity.

A study published in the July issue of Radiology found that digital CAD conducted with the Hologic R2 ImageChecker correctly marked 99 of 103 consecutive asymptomatic breast cancers, which is a 96.1 percent detection rate. Of those cancers marked, 44 were in the form of microcalcifications only, 23 cancers that appeared as a mass with microcalcifications and 32 of 36 lesions that appeared as masses only.

Those findings sync with Ruiz’s experience. “CAD is very good in spotting calcifications and discrete masses,” says Ruiz. “The weakest link in the CAD chain is architectural distortion, which is also true for human interpretation.”

For Melton, CAD comes in handy at the end of a long day when having another set of eyes—even computer-generated eyes—can help the radiologist pick up on something he or she might have missed. Melton has both digital and digital CAD images available from exams done at Columbia at his home office in West Hartford, Conn.; he has worked remotely for several years and finds that higher quality digital systems lead to a much-improved interpretive environment for the radiologist who has no comparison with the analog images used by more than 70 percent of imaging centers in the U.S.


CAD & economics

The cost of digital CAD can be a barrier to entry for some imaging centers and hospitals, especially those on the borderline of having enough patient volume to make a convincing cost/benefit case. Some studies suggest that it may take an imaging center as much as three years to recover the costs of investing in digital technology; Melton disputes these figures, stating that a center conducting 80 cases a day can be in the black within 9 months.

“If you look at the cost of film, chemicals and other costs associated with film-based mammography and then take into consideration the reimbursement for the scan and interpretation, you can break even after 90 days,” he says. “And CAD adds another $15 to $18 in reimbursement for each study, so that helps a facility break-even faster.”

The newer systems with more sophisticated algorithms will spur the adoption of CAD going forward, Ruiz believes. “I would not want to go back to the analog or pre-CAD days,” he says. “CAD has a lot to offer both highly expert readers and those who aren’t as experienced.”


Workflow improvements

Digital CAD is much easier to integrate into an efficiently designed radiological workflow than analog CAD, says Ruiz. Prior to digital CAD, films had to be fed into a digitizer; acquiring the CAD piece of the puzzle was more time-consuming, resulting in delays in interpreting images using CAD.

“Now with PACS, the CAD process is invisible to the technologist and the radiologist,” he says. “Now we can access CAD with a click of the mouse, or a stroke on the keyboard and you can view the CAD on your existing display.” Some radiologists like to overlay the CAD image with the non-CAD images for comparison purposes. Previous CAD views for a particular patient also can be compared side-by-side or by overlaying images, making it easier for radiologists to observe changes in breast tissue.

And because patients don’t have to wait for the technologist to confirm that the appropriate image was captured, radiologists are finding it easier to batch the interpretation of a number of images at once. CAD is easy to integrate into that process because it is so accessible and available on demand either at the location where the images are taken or at remote offices and home offices that have the necessary high-speed internet connections and high-resolution monitors as well as other equipment.

Melton’s experience interpreting thousands of images from his home office reveals that there is no difference in the image quality for both non-CAD and CAD digital mammography images that are read onsite versus those read remotely. Radiologists who desire to consult with colleagues not onsite can easily arrange remote access for those colleagues to both the digital images and the CAD images, potentially improving interpretation accuracy and efficiency for radiologists as well as reducing wait-time for patients.