Computer-aided detection (CAD) seems like an obvious next step in the evolution of image interpretation. Automation is taking root in so many fields, and radiology is no exception. A tool that can help spot lesions or other areas of interest would be incredibly valuable, even if the technology can’t replace a highly trained professional (and it’s not likely that most people would be comfortable turning over image interpretation to a machine, anyway).
But despite the potential, CAD has really only taken off in use with mammography. There are a few reasons for the seemingly slow rate of implementation. The first is financial: at the moment, reimbursement for using CAD is only available for mammography, leaving out many of the other possible applications.
Another reason, however, is that providers still may need to be convinced of everything CAD brings to the table. One of the top stories this month deals precisely with this issue, providing another example of the benefits and limitations of the technology.
Ji Hye Min, MD, of Sungkyunkwan University School of Medicine in Seoul, and colleagues sought to test the use of a CAD algorithm for the identification of coronary artery stenosis on coronary CT angiography (CTA). Their study, published in the April issue of the American Journal of Roentgenology, included 128 consecutive patients with acute chest pain in the emergency department (ED). Coronary CTA data were analyzed with a customized CAD algorithm that automatically segments the coronary artery tree, labels the major coronary arteries and detects stenotic lesions without human interaction.
The CAD algorithm yielded 100 percent sensitivity and 100 percent negative predictive value in per-patient analysis, however, a relatively high rate of false-positives pushed the specificity and positive predictive value down to 23.1 percent and 68.1 percent, respectively.
Min and colleagues concluded that while the high number of false-positives underscores the need for expert human radiologists, the high negative predictive value of the CAD tool would make it especially valuable as a “second reader” in an ED setting to speed discharge of patients with normal CTA results.
CAD may not be a magic bullet, but as more studies are conducted to better define the potential of the technology, the more it may be utilized in a wide array of clinical applications.
Editor – Health Imaging