Advanced Visualization for CT Lung Screening Creeps into Practice
Findings Navigator on visualization software identifies a pulmonary nodule of interest. Source: Siemens
Thoracic radiologists are at odds with one another about the benefits of image enhancement and analytical tools for low-dose CT lung cancer screening. Some consider advanced visualization software a timesaver that simplifies screenings by automatically locating and measuring suspicious nodules to determine their probable malignancy. Others say it lacks plug-and-play features for integration into their standard workflow and PACS, although some systems may work around that problem.

Some question the accuracy of measurements from automated software and the cost-effectiveness of CAD for lung cancer screening. Others say lung CAD improves efficiency and catches cancers they would otherwise miss. Instead of spurning CAD, they recommend it to every radiologist who performs CT lung screenings.

While disagreeing on some points, there is agreement on one issue; the lack of reimbursement for low-dose CT lung screening is the primary source of their problems. No major insurer covers CT lung screening, and there is no way for providers to bill for the cost of advanced visualization and CAD software that aids its interpretation.

A recent positive development is the introduction of the Lung Cancer Mortality Reduction Act of 2011 in the U.S. Senate (S. 752), which targets lung cancer screening and would fund a Lung Cancer Detection Program, designed to provide low-income, uninsured and underserved populations that are at high risk for lung cancer access to early detection services.

Likely game changer

The situation could change, however, this summer with publication of final results for the massive National Lung Screening Trial (NLST). Launched in 2002, the multicenter, randomized controlled trial is designed to measure the effect of low-dose CT screenings compared to chest x-ray on lung cancer mortality.

Final publication will expand upon preliminary findings announced in November 2010. It found that subjects who received at least three annual CT screening exams experienced a 20 percent drop in lung cancer mortality compared with patients undergoing chest radiography. Although that finding may sway some payors, most insurers are deferring judgment until next year when a highly anticipated cost-benefit analysis associated with the trial will be released, notes Ella Kazerooni, MD, director of cardiothoracic imaging at the University of Michigan and regional principal investigator for the NSLT.

Well-suited for visualization tools

Considering its complexity and low specificity, CT lung screening provides numerous opportunities for advanced visualization applications. Most pulmonary nodules detected with screening CT are less than 1 cm in diameter and provide little additional information about possible malignancy, says Kazerooni. Most are actually not cancerous, and those that are malignant will reveal their status by aggressive growth over time.

"This situation commits the patient to several CT exams over several years to determine if the nodules are growing," she says.

Visualization software can do more than just initially measure the location and sizes of suspicion lesions. They can be numbered, linked and stored for quick retrieval and automated comparative measurement from scan to scan, says Henry Krebs, MD, a radiologist at St. Joseph's Hospital in Atlanta.

Krebs uses Siemens Healthcare syngo.via workstation with lung CAD and nodule segmentation tools and credits nodule measurement automation for reducing his reading time by at least 25 percent.

"It saves us a good amount of time in doing direct side-by-side comparisons and providing accurate statistics to determine if there has been growth," he says.

To improve accuracy, Krebs has shifted away from routine reading of 5 mm axial views to 1 mm slices with overlaps processed through the workstation's 3D reconstruction engine and thin-client server.

Similar interpretative scenarios are possible with the TeraRecon Intuition platform, according to Krebs. A thin-client feature in both sets of software allows him to launch a specific study into the thin-client software while working on his PACS workstation. "We strongly believe that thin-client is the only way to do things," he stresses.

In fact, Krebs and his colleagues have yet to apply lung CAD software to all CT screening exams because of a cumbersome PACS interface.

The inability to integrate lung CAD with PACS also is a sore point with Kazerooni.  

"I may have five radiologists reading chest CTs on a given day. They can't all go to that one unintegrated device," she says.

To solve the problem, Kazerooni's department purchased advanced visualization software including ImageChecker CT lung CAD software, an option on Vital Image's Vitrea workstations. The entire package will be embedded in the medical center's PACS.

Relying on MIP

Radiologists at Roswell Park Cancer Institute in Buffalo, N.Y., switched from 5 mm to 0.25 mm slice analysis for reconstructed exams with the adoption 64-slice CT for their CT lung screening studies. The complete set of axial images is read in a cine mode on a GE Advantage Workstation.

Most tools applied at Roswell for detecting and tracking suspected nodules have been in use for at least a decade, says Alan Litwin, MD, co-director of CT and ultrasound. Litwin and Peter Loud, MD, director of body imaging, also rely on maximum intensity projection (MIP) imaging. It has proven to be especially good for evaluating peripheral nodules. They are investigating CT perfusion for mediastinal nodes, and use volumetric measurement to assess nodules that have irregular contours.

At MD Anderson Cancer Center in Houston, the standard screening protocol calls for the acquisition of 300 to 400 2.5 mm slices to assure lung coverage. Reginald Munden, MD, professor of radiology, stacks the reconstructed axial view in a cine mode with lung, soft tissue and bone windows on a Philips iSite workstation. Electronic calipers are used to measure nodule dimensions.

Munden has thus far resisted adopting advanced visualization tools because of the inefficiencies involved with physically moving from his accustomed position at the PACS to an independent workstation. They also tend to work poorly for measuring the large lung lesions that typically appear in his routine diagnostic practice.  

In his opinion, lung CAD does not yet perform well enough to serve as his second reader. It does not detect ground glass opacities, which warrant close attention in his department. Although CAD may detect small nodules that Munden misses, it is less adept at picking up large lesions, especially when they abut blood vessel, he says.

Over time, however, Munden expects to develop a synergistic relationship with advanced visualization and CAD, with the expectation that by working together man and machine will lead to high rates of detection and clinical success.

The transition to high-risk CT screening and integration of advanced visualization tools both remain works-in-progress. However, positive developments seem to be on the horizon.