Breast tissue complexity may predict false-positive recalls

Tuesday, December 2 | 3:10 p.m.-3:20 p.m. | SSJ01-02 | Arie Crown Theater
In this presentation, researchers from the University of Pennsylvania will describe how computer-extracted breast tissue complexity features may be able to decrease false-positive recalls in screening mammography.

The recall rate from screening digital mammography studies in the U.S. is approximately 10%, and the vast majority of these recalls are false positives. This contributes to an overall low positive predictive value for screening mammography and frequent overimaging in breast cancer screening, said presenter Shonket Ray, PhD.

Increased breast density is associated with a greater risk of a false-positive recall, and the researchers hypothesized that parenchymal texture patterns indicating a complex breast pattern could have a similar association.

"Ideally, both breast density and texture could be combined into an imaging index that accurately characterizes breast complexity to aid radiologists in mammographic interpretation and, potentially, in the decision-making on whether to recommend supplemental forms of breast cancer screening (i.e., tomosynthesis, ultrasound, or MRI)," Ray told "This type of tool could help further our ability to [provide] personalized breast cancer screening protocols."

Including parenchymal texture features improves the prediction of false-positive recalls compared with models that use only density measures, the researchers found. Their results also indicate that digital breast tomosynthesis (DBT) could be of potential benefit for the subset of women at high risk for false-positive recalls; breast tissue complexity was more strongly associated with a recall decision in patients screened with digital mammography alone compared with those who underwent DBT screening, Ray said.

"Therefore, this new model could help guide radiologists in selecting patients for DBT screening if tomosynthesis resources are limited," he said.

Page 1 of 155
Next Page