After systematically reviewing 14 articles encompassing eight high-quality observer/performance studies and 1,868 radiographs and CT scans, two influential researchers are recommending closer consideration of computer-aided diagnosis (CADx) for regular use in lung-cancer screening by general radiologists.
Guy Amir, MD, MPH, and Harold Lehmann, MD, PhD, both of Johns Hopkins University’s health sciences informatics division, compiled results showing consistently improved accuracy of supervised lung-CADx systems over traditional reads.
Their report posted Nov. 23 in Academic Radiology.
Concentrating on eight studies with high scores for quality assessment of diagnostic accuracy—all above 70 percent in QUADAS scoring—Amir and Lehmann noted “important improvement in accuracy afforded by CADx” to assessments by general radiologists.
The lone study that didn’t reflect significant improvement by CADx warranted a figurative asterisk, as it was supervised by thoracic specialists rather than general rads, according to the authors. Meanwhile this study did show significant improvement by CADx among a subgroup of resident radiologists (P < .009).
Imaging modalities included in the Amir-Lehmann review were radiography, low-dose CT and high-resolution CT.
In their discussion, the authors state the present review of the literature is unique in its focus on lung-CADx.
They also state that it has direct implications for other cancers.
“CADe, the precursor technology, has gained significant traction for mammography, achieving Food and Drug Administration approval in 1998,” they note before citing two reports indicating a 20 percent gain in early breast cancer detection.
(Those findings are at odds with a recent analysis of CAD in nearly 626,000 screening mammography exams that found the technique no better than traditional reads.)
“Although CADx is applicable to numerous organ systems, mammo-CADe followed by colon and lung have led current research efforts,” the authors point out. “The future is bright for CADx applications to other organ systems.”
They stress that their recommendation to consider including CADx in strategies for lung cancer screening studies and regular clinical workflow applies to general radiologists.
“Advances in accuracy must be achieved,” conclude Amir and Lehmann, “before these systems can significantly augment the performance of thoracic subspecialty-trained radiologists.”