Lung CAD: A Power Tool in the Imaging Toolset

	Siemens’ syngo lung CAD offers capabilities for the detection of solid pulmonary nodules in thoracic CT studies as well as CAD workflow enhancements.

Progressive clinicians are working with vendors to fine-tune CAD applications for lung imaging. Even as the tool can help radiologists detect more lung nodules, it’s only one item in a complete imaging toolset. But there’s a Catch-22: More radiologists need to use it to help prove its value, but ease of use has to improve to encourage more radiologists to use it.

Lung CAD is “not a shot in the dark or a flash in the pan,” says David Naidich, MD, professor of radiology and medicine at the New York University School of Medicine & Hospitals Center in New York City. He uses syngo Lung CAD from Siemens Medical Solutions. Lung CAD has been making “very steady progress,” he says.

Overall, lung CAD can aid imaging, decrease false negative findings, offers high sensitivity and has brought a 26 percent reduction in missed actionable nodules.

“As the technology got better and better in CT, generating huge datasets, the concept that computers would be an additional means of interpretation became very attractive,” Naidich says. That has turned out to be a very complex set of problems for the computer to solve.

Naidich doesn’t see CAD as an independent, stand-alone tool, even though it is designed to be just that. CAD isn’t used to detect nodules—which can determine whether a patient requires surgery—but also to characterize, conduct volumetric assessment, and look at contour and density to match scans later to see whether nodules have changed. Coupling CT and CAD allows for more accurate judgment in lung cancer cases, he says.

To facilitate a smooth workflow, study data are streamed to Siemens Leonardo workstations. But, Naidich hopes to soon see CAD integrated into PACS configurations. “When you work on two separate systems, you have to go back and forth. It would be easier if the tools were available right from the get-go.” That ease of use would translate, Naidich says, to physicians using CAD more often and more appreciation of its value. “Until [physicians] really are using CAD, they don’t fully sense that it’s a relatively simple idea. The information you’re getting is not onerous. As long as the device doesn’t overload you, it is truly positioned to only show you things you want to see for reassurance.”

Edwin van Beek, MD, PhD, radiologist at the University of Iowa Hospitals and Clinics in Iowa City, has been working with EDDA Technologies to boost the user friendliness of its IQQA Chest Enterprise software—requiring fewer clicks and offering more automation. The facility conducts research on lung nodules, so the partnership was a good fit.

Since implementing IQQA Chest Enterprise from EDDA Technologies 18 months ago, the software is used for annual follow-up on all lung cancer patients. The teaching facility is fortunate in that its residents already double-read all chest x-rays. A review found that of 214 cases, CAD was the only thing that caught a finding in 19 cases. Of those 19 cases, 16 were true positives and three were false positives.  Those are very good results considering that it takes about 20 seconds to run a chest x-ray through CAD.

A review of residents’ usage has shown that CAD helped them pick up 20 percent more nodules; for senior radiologists, that percentage is closer to 5 percent.


Skeptical interest

During a talk van Beek gave in June, he noticed a lot of skepticism but, at the same time, more interest in CAD than in recent years. Growth in the use of CAD depends on reimbursement, he says. Currently, there is a tracking code for CAD but “if there’s no reimbursement, people are not going to make the investment.” But, he feels that knowing that he’s done everything he can gives him peace of mind. The issue of missed or delayed diagnosis weighs on radiologists, however. “I have no problems sleeping at night, and I think a lot of people see it in a fairly similar way.”

Freiburg University Hospital in Hamburg, Germany, is one of the largest academic teaching hospitals in Europe with 1,600 beds and 1,150 physicians. The department of radiology employs 35 radiologists and 65 radiologic technologists, and an MR research section staffs 50 physicists. The team performs about 500 exams a day.

The facility implemented X-ray Lung Nodule Analysis (xLNA)—EDDA’s IQQA Chest Enterprise as marketed by Philips Medical Systems—in May, says Elmar Kotter, MD, senior radiologist. Although research has established CAD’s value with lung CT, its value with chest x-rays is still under evaluation, he says. Kotter has studied results from the use of xLNA and found encouraging results that he submitted for publication in Investigative Radiology.

CAD needs to be very fast and integrated into radiologists’ daily workflow, Kotter points out. He has found that xLNA is simple to integrate with PACS and is easy to learn.

Patricia Shapiro, MD, of Southcoast Medical Imaging, Savannah, Ga., has been using lung CAD from Riverain for about six months. After going all-digital years ago and implementing CAD for mammography almost two years ago, implementing lung CAD was pretty seamless, she says.

The facility implemented lung CAD because, as a multispecialty group, primary-care physicians read their own chest x-rays. “We were interested in doing this as a way to provide an additional tool to primary-care doctors.” While it has proved to be very helpful for primary-care providers, lung CAD hasn’t made as big an impact with radiologists yet, she says. “As lung CAD algorithms improve with time, it will make more difference to radiologists.” Meanwhile, CAD has value right now in her view if it encourages primary-care physicians to take a second and third look at chest x-rays.


In the future

Debate continues about whether CAD is a valid tool, Naidich says, and clinical evaluation has necessitated a large-scale, randomized, multi-facility study. “If the answer is that lung cancer screening is an important clinical tool in its own right, that would have tremendous impact on whether CAD to detect lung nodules would be something people want.” Study results are expected some time in 2008.