When to Use CAD in Breast Imaging
 | |
| Finn Lindhardt, MD, senior consultant at the Breast Center in Viborg, Denmark, discusses a mammography study with a colleague. Image courtesy of Siemens Healthcare. |
With increasing awareness of prevention and treatment options available to treat early-stage breast cancer, screening and detection aids are at the top of radiologists’ “must have” lists. However, challenges remain to the overall market expansion of CAD, as many in the clinical community remain unconvinced as to the true clinical value of radiology’s “second set of eyes” in detecting and diagnosing several types of carcinomas outside of mammography.
Mammography’s mainstay
While CAD technology for mammography continues to evolve and improve, it isn’t without a few stumbling blocks along the way. In 2007, mammography CAD took a hit with the publication of an article in the New England Journal of Medicine which claimed CAD lowered diagnostic accuracy due to the number of false positives it produces. As a result, some third-party payors contemplated reducing or cutting reimbursement. But CAD held fast, supported by a large body of scientific evidence that proves its value. Today, mammography CAD continues to detect up to 20 percent more cancers and helps detects nearly 75 percent of actionable missed cancers.
“CAD is an add-on for diagnostic work or in a screening situation, to help you find something that you might miss,” says Finn Lindhardt, MD, senior consultant, Breast Center in Viborg, Denmark, a user of syngo MammoCAD from Siemens Healthcare. “You can use it for dense lesions in the breast, so that the system can mark regions of interest, showing findings [that] are suspicious.”
CAD usage in mammography acts as a marker, urging the attention of a radiologist, either experienced or inexperienced, to lesions that might otherwise get overlooked or calcifications which could be the first sign of early breast cancer. However, Lindhardt says that CAD is just another add-on tool in the radiologist’s armamentarium to help in diagnosing and screening.
“If you are an experienced mammogram reader, the use of the CAD system is not much help. But for inexperienced readers, it is more beneficial since it will find more lesions,” he says, adding that what the technology still lacks is the ability to characterize lesions or calcifications. “In Europe, we don’t get reimbursed for using CAD, but if you decide to use it, you should use it on every case to be secure that you are choosing the right lesions for scrutinizing.”
Wynn W. Adams, medical director of Kettering Breast Evaluation Centers in Ohio, agrees that mammography CAD is an add-on tool, highlighting areas of suspicion for lesions 1 cm or less in size as well as calcifications. “We are finding more things to look at and question as we become more digital and iCAD’s SecondLook Digital CAD technology is helpful in analyzing areas of suspicion.”
For Adams, CAD usage is not a necessary component of mammography workflow. “You don’t have to use it, but it is occasionally helpful,” she says. “It makes you feel a little more secure, but often it isn’t really going to add very much more that a radiologist isn’t capable of seeing for himself or herself.”
Breast MRI CAD in the clinical routine
While CAD for mammography is more like a “spell-checker” for radiologists to flag potential findings that might get overlooked, CAD for breast MRI is slightly different, proving useful, not as an add-on tool, or “spell-checker,” but as an integral component in the clinical routine.
“It is not a second reader, or early detection tool—it serves more as a computer-aided evaluation and visualization tool,” says David R. Gruen, director of the Connecticut Breast Center at Norwalk Radiology in Connecticut, and assistant chief of medical staff at Norwalk Hospital, who uses CADstream breast MRI CAD from Confirma.
What matters in breast MRI interpretation are both the morphology of a lesion, its shape, and the kinetics of the lesion, how it accumulates the intravenous gadolinium and subsequently washes out. “The reason this is important is that malignancies tend to take up intravenous contrast rapidly and then tend to wash out the contrast rapidly as well. So when you put this together—abnormal morphology and suspicious (type III) kinetics—you start to get a reasonable positive predictive value for something you might need to biopsy,” Gruen says.
What CAD brings to this mix is a reasonable way to organize the data, since the software is good at locating suspicious areas and identifying whether or not they enhance, and if so, measuring whether they enhance in a benign or suspicious manner.
“All of the features that might suggest malignancy are at your fingertips,” he says. “Most bilateral breast MR exams consist of 2,000 to 3,000 images. If you have to evaluate all this data manually, and do it multiple times a day, you could easily overlook something, measure the wrong lesion, look at the enhancement kinetics of the wrong lesion and miss or misdiagnose an early or subtle breast cancer. This is why it really is a necessary part of the interpretation process both to find and to characterize subtle lesions, not as a second reader, but as an essential part of the interpretation process.”
“Using CAD can decrease false positive rates and subsequently the number of unnecessary biopsies,” Gruen points out. “Looking forward, the key challenge is to continue to increase specificity without sacrificing the exquisite sensitivity of breast MRI. As an imaging test, it has the ability to alter the course of therapy for many women, and potentially save many, many lives.”
This lack of specificity is part of the reason that some imaging centers do not perform breast MRI, according to Rebecca G. Stough, MD, director of imaging for Mercy Women’s Center and clinical director for Breast MRI of Oklahoma in Oklahoma City. “The sensitivity is wonderful, but the specificity is a problem,” she says, adding that CAD improves breast MRI’s specificity. Stough is using AuroraCAD breast CAD from Aurora Imaging Technology, which is integrated with the Aurora 1.5T Dedicated Breast MRI system.
For example, if an enhancing mass with washout is found in the upper outer quadrant of a breast, Stough says she uses multiplanar reconstruction capability of CAD, does an oblique subtracted MIP reconstructed image, rotates that to find a fatty highlight and looking at the mammogram, can identify a lymph node and potentially rule it out—even with a washout curve.
Stough says she does not believe it is possible to interpret a breast MRI without the use of some kind of CAD. “If you don’t have CAD to assist in interpretation your ability for an accurate diagnosis is going to be limited,” she concludes.