In Review - News from the 2005 Meeting of the RSNA

Digital x-ray challenges CT lung cancer screening

Innovative x-ray technologies bolster early detection and shorten disparity between DR and CT

By: H.A. Abella

Low-dose CT appears well-suited to the task of diagnosing lung cancer at an early stage. But it faces challenges regarding small nodule detection, and it may not be as widely available for screening. Several research groups suggest that digital radiography-further enhanced by computer-aided detection and digital subtraction techniques-could overcome CT's shortcomings in this setting.

Li Fan, Ph.D., and colleagues at the Peking Union Medical College in Beijing found decreased interobserver variation and an increased nodule detection rate when CAD was added to DR. In 200 screening studies, the average detection rate without CAD was 50% per individual reader, but the collective detection rate averaged 35%, underscoring an important interobserver variation, Fan said.

With CAD, individual and collective detection rates each averaged 90%. The CAD program can be applied to computed radiography as well.

Overlying bone structures challenge DR's ability to detect lung nodules, but dual-energy subtraction chest radiography shows promise. Dr. Seong Kyu Ahn and colleagues in Daejeon, Korea, found that dual-energy subtraction eliminated rib shadows and allowed accurate, computerized measurement of lung nodule volume. In addition, Dr. Amy E. Musk and colleagues from the University of Maryland reported that using dual-energy subtraction radiography helped reduce the recommendation rate for follow-up CT.

Dr. Nariman Majdi-Nasab and colleagues at Duke University assessed automated DR-based tomosynthesis for lung nodule detection. They developed a CAD algorithm to identify nodules in five patients enrolled prospectively in an ongoing trial. The system performs image filtration, segmentation, and feature extraction while patients undergo tomographic imaging in 69 planes at 5-mm intervals.

The CAD/tomographic method detected 11 nodules and missed one, for an average sensitivity of 92%. The program marked an average of seven false positives per subject, primarily due to end-on vessels seen in cross section, Majdi-Nasab said. Researchers concluded that the system's sensitivity and specificity match those of other CAD methods used to project chest radiographs. The team is investigating ways to reduce false positives.

DR/computed radiography might be less expensive compared with CT. In reality, however, only a handful of institutions can offer the techniques proposed, said Dr. Cornelia Schaefer-Prokop, an associate professor of radiology at the Academic Medical Center in Amsterdam.

"A lung cancer screening program must consider four important aspects: availability of diagnostic imaging modality used; radiation dose; diagnostic performance; and overall costs involved in screening. Considering that low-dose CT is already a powerful and available tool, DR/CR studies must provide convincing evidence of that technology's superiority in these four areas," she said.