Study: CAD finds missed cancers, but boosts recall rates
Computer aided detection (CAD) increases the breast cancer detection rate and decreases the false negative rate, according to a prospective study appearing in a recent issue of the American Journal of Roentgenology. The drawback, however, is that it can lead to heightened recall rates.

For the study, radiologists evaluated 5,016 screening mammograms and detected 43 of the 48 cancers without the use of CAD, said Priscilla Slanetz, MD, Boston University Medical Center, an author of the study. With the addition of CAD, two additional cancers – both ductal carcinoma in situ that presented as microcalcifications — were detected, she said.

Without CAD, radiologist detected 90 percent of malignant asymmetry or masses and 89 percent of microcalcifications. CAD, however, marked 67 percent of asymmetry or masses and 100 percent of microcalcifications.

“CAD is best suited to detect microcalcifications, and in reality faint microcalcifications are the most commonly overlooked finding by most radiologists,” Slanetz said. Neither the radiologist or CAD was perfect, but the false negative rate was reduced when both were employed, she said.

As for recall rate, CAD resulted in an additional 89 women (from 607 to 696 patients) being recalled. Of those, six women were biopsied; and two had cancers, Slanetz said.

'”We continue to routinely employ CAD in the interpretation of screening mammograms. The tradeoff of earlier detection for a slight increase in recall rates seems reasonable and is well accepted by most patients,” she added.

Slanetz believes CAD is a useful addition to any practice, particularly in light of the ongoing shortage of breast imagers in the U.S. Radiologists need to be aware of what CAD can detect well, and what it tends to miss, she added.

In another study recently published in the same journal, CAD was found to be a potentially useful adjunct to double reading of screening mammograms.

The study included 3,683 women who had both film and digital mammograms which were then independently interpreted by two radiologists. “Twenty nine cancers were prospectively diagnosed in the initial screening round,” said Per Skaane, MD, of Ullevaal University Hospital in Norway, and lead author of the study.

Double reading of screen-film mammograms resulted in the diagnosis of 27 of 29 cases of cancer. CAD correctly marked 27 cases, of which one was missed at double reading.

“Double reading of full-field digital mammograms resulted in the diagnoses of 22 of the 29 cases of cancer,” said Skaane. “The CAD system correctly marked 27 cases and, of importance, marked all seven cancers missed in double reading of the full-field digital mammography cases,” he said.

Skanne added that “since CAD has the potential to mark tumors that might be overlooked by radiologists, adding CAD to double reading may mean these cancers could be detected at an earlier stage.”