Does access to FFDM priors affect DBT interpretation?

2015 03 25 11 32 00 880 Hakim Christiane 175

Does having access to full-field digital mammography (FFDM) priors improve radiologists' interpretation of digital breast tomosynthesis (DBT) images? Both exams contribute, but not in the same manner, according to a new study published online in Radiology.

The availability of prior FFDM and DBT images are each largely independent contributing factors in reducing recall recommendations during mammographic interpretation, according to researchers from Magee-Womens Hospital of UPMC. In addition, if priors are unavailable, DBT images alone are as good as -- if not better than -- the combination of current and prior FFDM images in terms of improving a radiologist's sensitivity.

"We wanted to explore how the impact of one type of study versus another affected the decision of whether a woman needs to be recalled or not," lead author Dr. Christiane Hakim told

Dr. Christiane Hakim from Magee-Womens Hospital.Dr. Christiane Hakim from Magee-Womens Hospital.

For the study, Hakim and colleagues asked eight radiologists to interpret 459 mammograms from 153 women. Each case included current FFDM, prior FFDM, and current DBT images that had been acquired at Magee-Womens Hospital between June 2009 and January 2013; of the 153 cases, 50 were malignant, 60 were negative and benign (not recalled), and 43 were benign (recalled).

Among the malignant cases, 16 were ductal carcinoma in situ, 31 were invasive ductal carcinoma, two were invasive lobular carcinoma, and one was a papillary carcinoma. Of the 153 women, 5.2% had fatty breast tissue, 28.1% had scattered fibroglandular dense tissue, 62.1% had heterogeneously dense tissue, and 4.6% had extremely dense tissue (Radiology, March 13, 2015).

The interpreting radiologists read the images in two different sequences, Hakim said.

"In the first sequence, the reader was shown an FFDM exam and asked to decide whether it required recall," she said. "If it was recalled, they were asked the likelihood of malignancy. Then, in that same sitting, the radiologist was shown the tomosynthesis exam for that case, and asked to evaluate again. Finally, he or she was shown the FFDM priors for the case."

"In the second reading sequence, our readers were first shown the FFDM exam, then the FFDM priors, and then the DBT exam, evaluating after each study," she explained.

The use of prior FFDM images reduced average recall rates for benign cases by 34%, and the combination of prior FFDM and DBT reduced recall rates for noncancer cases by 32%, Hakim and colleagues found. However, this reduction in recall rate came at a cost: a 7% reduction in sensitivity for prior FFDM exams alone and 4% reduction in sensitivity for the combination of prior FFDM exams plus DBT.

The use of DBT images resulted in less of a reduction in recall rate for noncancer cases: 19% without the addition of FFDM priors and 26% with them. However, sensitivity increased by 4% and 8%, respectively, with the availability of DBT.

"While availability of prior [FFDM] images resulted in a larger decrease in recall rates, it also reduced sensitivity, whereas availability of DBT images simultaneously decreased recall rate and increased sensitivity," the authors wrote.

The fact that the improvement in sensitivity associated with DBT remained stable -- whether or not prior FFDM images were available -- is good news for women who present for screening mammography without priors, perhaps because they're getting their first mammogram or because their prior mammograms were lost, according to Hakim and colleagues.

They plan to continue this research by investigating why the sensitivity went down in the combination of FFDM plus priors, she told

"Going forward, we're further exploring the question of what might have caused the sensitivity decrease when priors were available to the interpreter," she said.

Page 1 of 568
Next Page