Cancer-detection software proves a worthy, albeit imperfect, ‘second reader’ of lung CT

Computer-aided detection (CAD) may be useful as a “second reader” of low-dose CT lung images, as four CAD systems in a small study found up to 70 percent of lung cancers a radiologist had missed yet missed around 20 percent of cancers that human eyes had caught.

Corresponding author David Yankelevitz, MD, Icahn School of Medicine at Mount Sinai, and colleagues report their findings in the October edition of Radiology.

The team identified 50 lung cancers that manifested as a solid nodule on CT scans in annual rounds of screening and that could, in retrospect, be identified on the previous CT scans.

The team determined, by consensus of two radiologists, the total number of accepted CAD-system–detected nodules. They also recorded the number of CAD-system–detected nodules that were rejected by the radiologists.

They found that the four CAD systems detected from 56 percent (28 of 50) to 70 percent (35 of 50) of lung cancers in the earlier round of screening when the cancers were missed by radiologists.

False-positive findings varied from less than one per cancer to more than seven per cancer one year before the cancers were identified by the radiologists.

Meanwhile, the CAD systems missed about 21 percent (between nine and 13 of 50) of cancers correctly identified by the radiologists.

Of note, the team found clear differences in performance of the CAD systems. Some attained relatively high sensitivity for both nodule and cancer detection with a low rate of false-positive findings, while others, the authors write, “more heavily traded their increased sensitivity with the consequence of higher rates of false-positive findings.”

From this Yankelevitz et al. conclude that user preference may help cancer teams refine their choices of CAD software based on priorities for features.

Their findings also “suggest limitations for CAD systems as a primary or possibly even concurrent reader because they missed too many cancers that were detected by radiologists,” the authors conclude. “[F]or now, the most appropriate role of CAD is to use it as a second reader where the capability of the CAD system to detect at least some of those missed cancers is compelling.”

Dave Pearson

Dave P. has worked in journalism, marketing and public relations for more than 30 years, frequently concentrating on hospitals, healthcare technology and Catholic communications. He has also specialized in fundraising communications, ghostwriting for CEOs of local, national and global charities, nonprofits and foundations.

Trimed Popup
Trimed Popup