CAD system shows potential in detecting subtle lung cancers

A commercially available computer-aided diagnosis (CAD) system has shown promise in its ability to mark hard-to-detect lung nodules in standard and portable chest x-rays, according to study results published in the April issue of Academic Radiology.

Past observational performance studies have shown that radiologists miss approximately 40 percent of subtle cancers when reviewing conventional chest radiographs (CXRs), a fact that has spurred recent innovations in chest imaging technologies such as dual-energy subtraction, temporal subtraction and, specifically, CAD systems.

For the current study, authors Feng Li, MD, PhD, and his colleagues from The University of Chicago evaluated Riverain’s ClearRead +Detect system. “This CAD system recently incorporated bone suppression imaging (BSI) which, in addition to impacting CAD system performance, can improve radiologists' accuracy in the detection of lung nodules and discrimination between true-positive and [false positive] CAD marks on CXRs and through direct visualization of the BSI images.”

Li and his team set out to determine the effectiveness of the CAD system at detecting lung nodules in a large consecutive group of CXRs with CT correlation. To do so, they performed a retrospective analysis of 586 patients from the researchers’ medical facility who underwent either standard or portable chest radiography and CT scanning on the same day. The CAD system was then applied to all 586 CXRs, with the results cross-referenced against the corresponding CT images to determine true nodules and false positives.

Their results showed that in the consecutive series of CXRs, the CAD system marked 47 of 66 lung nodules (71 percent), with a mean false-positive rate of 1.3 per image across all 586 CXRs (1.5 per image on 438 abnormal CXRs and 0.8 per image on 148 normal CXRs). Non-nodule pathologic findings accounted for a total of 41 percent of the 752 false positives marks.

“For cases with multiple nodules, the ability of a radiologist to detect at least one of the nodules may lead to a CT scan, which would then reveal any additional nodules that may be present, and in this context, the CAD sensitivity was found to be 84 percent on a per-CXR basis,” Li and his colleagues wrote. “Importantly, when applied to the seven CXRs with a solitary nodule that was missed during clinical interpretation, the CAD system marked four of the seven nodules (57 percent); these four nodules included the only two primary lung cancers in this database that were not mentioned in the clinical radiology report.”

The authors say their results highlight the potential value of CAD systems for diagnosing subtle lung cancers, especially with regard to routine use on CXRs prior to CT imaging procedures.

“Although CT is likely to remain the preferred technique for lung cancer screening due to its greater sensitivity, many lung cancers are detected initially on CXRs that are performed in patients with nonspecific symptoms,” wrote Li et al. “Such cancers are often unreported, but visible on radiographs in retrospect. Routine use of CAD has the potential to direct attention to the suspicious findings in such cases, and thereby reduce oversight errors.”

John Hocter,

Digital Editor

With nearly a decade of experience in print and digital publishing, John serves as Content Marketing Manager. His professional skill set includes feature writing, content marketing and social media strategy. A graduate of The Ohio State University, John enjoys spending time with his wife and daughter, along with a number of surprisingly mischievous indoor cacti.

Around the web

The new technology shows early potential to make a significant impact on imaging workflows and patient care. 

Richard Heller III, MD, RSNA board member and senior VP of policy at Radiology Partners, offers an overview of policies in Congress that are directly impacting imaging.
 

The two companies aim to improve patient access to high-quality MRI scans by combining their artificial intelligence capabilities.