Integrating Lung CAD and PACS Streamlining Day-to-Day Usability
Philips’ xLNA CAD highlights three suspicious areas in a chest x-ray.
Fueled by promising clinical studies, computer-assisted detection (CAD) technology continues to improve early lung cancer detection: greater sensitivity, fewer false-positives. As developers work to refine CAD’s capabilities so as not to bog down image interpretation and accuracy, many radiologists are already successfully integrating lung CAD into day-to-day clinical practice, thanks to links with PACS. The future appears promising and we can anticipate a day when lung CAD moves from a “detection tool to a diagnostic tool.”

Since lung CAD’s debut on the market five years ago, clinicians and manufacturers have worked to refine and improve the technology to reach a level of sophistication that will speed CAD’s utilization to an everyday clinical tool that provides a greater level of clinical assuredness than previously available. Better and more efficient diagnosis of early stage lung cancer via CT and digital x-ray, that is what CAD is striving for. As refinements continue to be made, and clinical studies validate its effectiveness, lung CAD is moving its way into a part of the clinical routine for helping radiologists detect solid lung nodules earlier.

Integration into everyday workflow

For a new technology to be readily adopted into everyday practice, it is critical that it seamlessly blends into the existing clinical workflow, ?with minimal to non-existent disruptions.

At the department of radiology at RWTH Aachen University in Aachen, Germany, Siemens’ syngo Lung CAD is integrated into the PACS workstations, giving radiologists a second reader for interpreting complex chest CT data sets. It also aids in the detection of solid pulmonary nodules in thoracic CT studies.

The department also uses the CAD Server from Siemens. As long as a patient is not in the ICU, all thoracic CTs are sent to the server and CAD is used on the PACS workstation, according to Marco Das, MD, fellow in the department.

“We get preprocessed results from the CAD Server displayed in the PACS,” says Das. “We can then do direct comparison to the original data set and see if we missed anything.”

Prior to pairing Siemens’ syngo Lung CAD and server, CAD results had to be processed on a separate workstation, which required radiologists to interrupt workflow to move to a different workstation to process results. “Now with the CAD server and syngo Lung CAD, we have a solution that allows us to stay at the PACS station and have CAD results directly there to use for routine reading. This results in a decrease in image reading time and smoother workflow because the radiologist stays in one location.

Around the globe in Japan, CAD also is finding its way into the clinical routines of radiologists at Iwate Prefectural Central Hospital in Moroika.

For the last 10 months, radiologists have been using the xLNA Enterprise CAD from Philips (the OEM version of EDDA’s IQQA-Chest system) for visualizing, identifying, evaluating, and reporting of pulmonary lesions in digital chest x-rays. They’re benefitting from an increase in the diagnostic accuracy of detecting pulmonary nodules approximately 5 to 15 mm.

CAD systems are not perfect, but if they can be integrated into diagnostic workflow, it can be helpful in diagnosis, says Yasuo Sasaki, MD, PhD, director of radiology.

CAD, which is integrated into the hospital’s Kodak Carestream PACS Client Suite, has helped to streamline workflow. The software automatically loads when the chest image is selected and with a fast processing speed, “there seems to be no issue using the xLNA in daily clinical use,” Sasaki says, adding that staff use lung CAD on all chest x-rays taken in the hospital.

“As a doctor, it is always good to ask a second opinion, and we are able to get that with CAD,” Sasaki says.

Earning its spot, fueled by study results

Despite the significant advances made in imaging technology over the last few decades, the early detection of lung cancer remains a challenge. Neither the technologies nor the interpreters have reached the level of accuracy to make it an exact science, says Moulay Meziane, MD, section head of thoracic imaging at the Imaging Institute within the Cleveland Clinic.

Meziane has been evaluating how chest x-ray CAD can improve practical, early detection of lung cancer, as part of collaboration between Riverain Medical and Cleveland Clinic. Investigators conducted retrospective studies to evaluate the performance of the CAD system and the readers using it. The multiple-reader, multiple-cases research enlisted six expert chest radiologists, six general radiologists, and six pulmonologists to assess and confirm the benefits of using chest x-ray CAD.

Each physician was tasked with reading 200 chest x-rays on patients at high risk for developing lung cancer, of whom about 100 had cancer and 100 did not. All cases were CT confirmed, and all cancers were confirmed through either a biopsy or surgery. Riverain’s OnGuard CAD system was used to identify cancers that ranged in size from 7 mm to 29 mm.

“We wanted to evaluate the level of CAD performance based on our own patient population, and to evaluate reader performance,” Meziane says. “We found that CAD had the potential to improve the level of performance among all groups of readers, even among the most experienced and least experienced chest radiographers.”

Like similar research studies, the researchers found that less experienced readers improve their cancer detection rate by using CAD, but at the same time were adversely influenced by CAD false-positive markings. Such findings reinforce the need for proper training on how to efficiently use CAD, he adds.

Meziane and colleagues, who presented study results at the 2008 annual meeting of the Radiological Society of North America (RSNA), tested four subsequent versions of the OnGuard system, from its debut as OnGuard 1.1 through version 3.0, 4.0 and 5.0.

In their evaluation, they noted a gradual increase in sensitivity in each subsequent CAD version: OnGuard sensitivity was initially 44.2 percent; version 3.0 and 4.0 had a sensitivity of 62.5 percent; and the latest 5.0 version had a sensitivity of 64.4 percent.

While sensitivity increased, the researchers also found a reduction in the number of false positives per image with CAD—decreasing from an initial average of 3.9 to 2.0 false positives markings per images in the latest version.

“I cannot imagine CAD reaching a 100 percent accuracy in the very near future [no false positives or post negative in any given image], but CAD may become a clinically useful tool when it reaches a 70 percent sensitivity with 2 or less false positives per image,” Meziane says.

In the sample of 104 cancers, one third of the cancers had very subtle lesions which where hard to see by the interpreter. In another one-third of the cases, the lesions were moderately subtle lesions (those that could be found by CAD and potentially missed) and in the (final) third of the cases, the cancers were obvious.

“I do not expect CAD to significantly impact the detection of the very subtle or obvious lesions, but I do expect CAD to make a difference in the detection of the moderately subtle lesions,” Meziane says.

Looking to the future

Chest CAD will continue to experience some growing pains before seeing increased widespread clinical acceptance within radiology. One hurdle is the number of false positives. Once that diminishes, and sensitivity increases, trust will be built for the technology. Das of RWTH Aachen University says that one area of future research would be to focus on CAD’s clinical impact on outcomes. “Since we have evaluated the software under experimental settings, we need to look at [CAD’s] clinical outcomes: what impact does CAD have in changing the clinical outcomes of patients with lung cancer?”

Meziane also is participating in a three-year prospective trial funded through a grant from the state of Ohio, Department of Defense, to evaluate the performance of lung CAD and its readers. Recruitment of 8,000 enrollees began in January 2009 to examine how CAD benefits the outcome of lung cancer, using OnGuard 5.0.

This trial is the first in the Early Lung Disease Detection Alliance (ELDDA), an ongoing research and commercialization program that will develop, test and bring to market new image-analysis systems for the early detection of lung cancer and other lung diseases.

“Ultimately, this is a product that will be integrated in our common practice and the time will come when it will reach the level of accuracy that will allow it to become a reliable diagnostic tool,” Meziane concludes.  “More research will gradually improve the CAD products, and it will be just a matter of time before CAD will transition from a detection tool to a diagnostic tool.”