Dear AuntMinnie Insider,
Computer-aided detection (CAD) remains a key driver of innovation in virtual colonoscopy, thanks to increasingly sophisticated research aimed at minimizing false-positive findings while highlighting true polyps that might otherwise be missed.
Toward that end, researchers from the University of Chicago achieved impressive results with an algorithm based on massive-training artificial neural networks (MTANNs), which helped even experienced readers find more true polyps.
The study was limited to patients with lesions that were missed in a large U.S. trial. Get the details of our story by AuntMinnie.com senior editor Erik L. Ridley in our Insider Exclusive, brought to you before it appears on our general site.
On the human side of polyp detection, many studies have shown the value of dedicated training on real VC data for achieving the high accuracy of an experienced reader. But matters get fuzzier when it comes to exactly how many training cases are needed. That's why a new study from Erasmus Medical Center in Amsterdam is so intriguing. Marjolein Leidenbaum, MD, and colleagues looked at multiple readers and multiple sets of VC data to arrive at an answer that you'll find by clicking here.
Optimal colonic distention also is important for polyp detection, of course, and there's more than one way to achieve it, say researchers from Wisconsin. Their study found that a third scan series with the patient in a decubitus position rendered many datasets with suboptimally distended colonic segments more readable.
Finally, there's nothing quite like adding a second valuable screening exam to render a screening test such as virtual colonoscopy more useful and cost-effective. A new study confirms the value of VC data in the evaluation of bone mineral density, a vastly underdiagnosed condition that leads to significant costs and health problems when left unaddressed. Click here for the details, and be sure to scroll down for all the news that's fit to click in your Virtual Colonoscopy Digital Community.
