Lung CAD Carves a Niche
Finding lung cancer at its earliest, and thus most treatable, stages is the objective of lung CAD. Advancements in CAD's detection and greater workflow friendliness are helping to boost interest. Clinical sites are running the lung CAD systems that review CT or x-ray images through their paces, while a few vendors are set to launch new options in the next several months.
Since first hitting the market in 2001, lung CAD has been garnering support in the radiology community. Lung CAD systems are available to aid in the detection of suspicious nodules on both x-ray and CT studies. Both types of solutions are showing that they can help detect lung cancer in its earliest and most treatable stages. Some systems incorporate tools that simplify follow-up of suspicious and malignant lesions, such as automatically measuring nodule volume or registering previous scans. The combination of enhanced detection and workflow gains translate into some bold predictions.
"Lung CAD is going to become part of the day-to-day workflow," predicts David Yankelevitz, MD, professor of radiology at Weill Medical College of Cornell University. "Relatively soon, radiologists will be using this technology routinely."
But users agree that lung CAD technology has not yet been perfected, and vendors are working to further boost the utility of the technology. Many are signing partnerships with PACS and 3D vendors that outline plans to integrate CAD into those products. These agreements will facilitate integration and workflow and help lung CAD establish a firm presence in the radiology department.
The chest x-ray route
One option for adding CAD to the lung cancer detection arsenal is to use an x-ray-based CAD system. Supporters of the approach tout the widespread use and low cost of chest x-rays.
Garden City Hospital in Garden City, Mich., has analyzed approximately 700 chest x-rays with Riverain Medical's (formerly Deus Technology) RapidScreen CAD system since installing the system a year ago. The hospital provides CAD free of charge to all chest x-ray patients meeting one of four criteria: smoker, second-hand smoke exposure, asbestos exposure or factory employment.
Analog chest films are digitized and pushed through RapidScreen software. Radiologists typically read the film and CAD print outs simultaneously. Radiology Administrator Jim Williamson notes, "It's important to create a process for reading films and CAD printouts." This links CAD and film results and keeps workflow moving. Williamson says workflow issues are minimal, with CAD adding a maximum of 30 seconds to each exam.
The majority of CAD marks are false positives as the software does pick up tuberculosis scars and rib junctions, which can be readily dismissed by a radiologist. RapidScreen also has picked up one 4 millimeter mass missed by a resident, enabling clinicians to treat a cancer that would have become lethal within months.
For those in the digital realm, Riverain markets RS-Digital for use with digital x-rays, and GE Healthcare has integrated the CAD system into its Revolution DR and CR systems. Both options can process dual-energy subtraction images, which optimizes viewing by reducing obstructions and separating bone and soft tissue images.
Other lung CAD systems focus on CT scans, but the systems are not mutually exclusive as findings on chest x-rays are often followed by CT scans.
The CT option
One of the newest CT CAD options is Medicsight's Lung CAR system. Cornell's Yankelevitz says, "The system performs quite well, finding small lung nodules in the four to five millimeter range, which are often overlooked and the range that you don't want to miss. It is also not missing nodules that turn out to be cancers."
Lung CAR incorporates an adjustability feature that allows sites to fine-tune the system's sensitivity. For example, on an initial screen the radiologist can focus on five to eight millimeter nodules and adjust the "highest sensitivity sweet spot" down on follow-up studies. Or if the radiologist is searching for metastatic diseases in a thyroid cancer patient, the dial can be set for the very small nodules indicative of metastasis. Users also can enhance performance by adjusting slice thickness, dose and speed. Yankelevitz says the adjustability feature is critical in optimizing CAD.
Medicsight has signed agreements with Agfa, Viatronix and Vital Images, outlining plans to integrate Lung CAR into their systems.
R2's ImageChecker CT Lung is another CT CAD option. San Francisco General Hospital has relied on it as a second reader tool to enhance detection since the product's research and development phase several years ago. "We filter cases for follow-up, detect lung cancers and use it to aid in the detection of metastatic disease in the chest," explains Michael Gotway, MD, acting chief of the radiology department.
The one technical prerequisite for ImageChecker CT is a multislice CT scanner as the technology is designed to work with collimated slices under five millimeters. The system is networked into PACS, and radiologists read the CAD studies from the ImageChecker workstation. Switching back and forth between PACS and CAD workstations can create minor workflow issues, says Gotway. R2 is working on integrating the system into several vendors' PACS to overcome the issue.
"Radiologists make up any lost time with the temporal comparisons," Gotway continues. The chest CAD system lines up two CT scans, automatically detects nodules for comparison and creates an instant report. This can cut reading time in half in severe cases, Gotway estimates. Moreover, automated reports contain a high level of detail, which can aid referrals.
False positives remain a challenge, as radiologists do need to weed out marks that indicate normal features such as vessel crossings or scars. But most CAD and second reader systems average three to five marks per case, which is quite manageable. Another minor challenge is CT protocols; hospitals may need to adjust protocols to reconstruct chest CT studies at five millimeter or thinner slices, and they must decide which patient populations are most appropriate for CT CAD.
ImageChecker CT Lung also is marketed by Vital Images as an option on its Vitrea 2 workstation. Edison Imaging Associates in Edison, N.J., has relied on the software for nearly a year. Vitrea 2 workstations are networked into the center's PACS, allowing radiologists to use the software to look for nodules and measure volumes on 16- and 64-slice CT datasets from referring locations around the state.
R2 recently signed an agreement with Kodak to integrate ImageChecker CT in its PACS workstation.
Other options
Lung CAD isn't the only post-processing technology to facilitate lung cancer follow-up. Siemens Medical Solutions offers syngo LungCare, a set of tools for viewing lung nodules that includes Nodule Enhanced Viewing (NEV), a second reader tool. Like CT CAD systems, the radiologist must read through the case before the software highlights marks for further analysis. The biggest bang for radiologists comes in the form of the analytical tools that simplify and streamline evaluation of nodules, says David Naidich, professor of radiology at New York University Medical Center. This includes maximum intensity projections, volumetric projections, 180-degree rotations, instant 3D segmentations and automatic registration of follow-up scans. The automatic registration translates into healthy workflow savings, as comparing two CT scans on a point-by-point basis can be a time-consuming process.
The FDA cleared GE Healthcare's Advanced Lung Analysis (ALA) software in June 2004. The technology creates 3D models of lung nodules and automatically measures them in volume. "This is a more accurate and objective way of determining the size of nodules," explains, Robert Klym, MD, radiologist with Pardee Hospital in Hendersonville, N.C.
The only technical prerequisite for ALA is a multislice scanner that scans nodules at 1.25 to 2.5 millimeters or a reconstruction program to reconstruct nodules in that size range.
There are minor technical and human resource challenges, says Klym. Sometimes the radiologist will need to override the software as the program can have a hard time defining nodules adjacent to some structures. Most CAD systems also experience some difficulties with nodules close to certain structures and very small nodules.
The other thread among users of all types of systems is educating radiologists so that they understand the value of the tool and realize it won't usurp their position.
The ideal lung CAD
"The gold standard is a CAD system integrated into PACS that incorporates temporal comparison and provides accurate detection with a minimum of false positives. We're not quite there [yet]," sums Gotway. Naidich adds, "The biggest problem with lung CAD is that studies must be read on an independent workstation. PACS integration will be the next really big step."
The CAD market is still maturing, with new products poised to hit the pavement in the next year. Those already in the market are still gearing up, angling for partnerships with PACS vendors. This path may be the best way for CAD companies to get their products integrated into radiology departments, concludes Yankelevitz. And hospitals also will gain as PACS integration will enable them to configure CAD tools to meet individual user's needs. Stay tuned.
ON THE HORIZON
Radiologists and lung specialists can bet on the lung CAD market becoming more crowded in the coming months.
EDDA Technology (www.eddatech.com) recently released its IQQA-Chest soft-copy chest CAD product. The system is designed to aid in the detection and quantification of five to 15 millimeter lung nodules and lesions in digital radiographic chest images. It offers a three-task approach, with an image reading page with image visualization tools, a region of interest (ROI) analysis page with automatic segmentation and measurements and clinical report page that automatically generates a report.
iCAD (www.icadmed.com) is developing a plug-in module to work with third-party 3D reconstruction software. The system will accommodate thin- and thick-slice scanners, including single slice and spiral CT systems. Currently in field trials, the CAD system may be released by the end of 2005.
Median Technologies' (www.mediantechnologies.com) CAD-Lung is a works-in-progress CAD application that provides the radiologist with tools to support the identification and characterization of lesions and evaluate of lesion evolution. According to the company, the product can be integrated with HIS, PACS and review workstations. CAD-Lung will be submitted to the FDA as a 510(k) early this year.