AI speeds up DBT reading time, helps find more cancers

Monday, November 26 | 11:20 a.m.-11:30 a.m. | RC215-14 | Arie Crown Theater

In this talk, researchers will report that the concurrent use of artificial intelligence (AI) software while interpreting digital breast tomosynthesis (DBT) screening exams leads to higher radiologist accuracy and much faster reading times.

Adding DBT to full-field digital mammography has been shown to lower recall rates and improve cancer detection; however, it takes significantly longer to read a DBT study, according to presenter Dr. Emily Conant of the Hospital of the University of Pennsylvania in Philadelphia. The researchers sought to determine whether AI software from iCAD -- currently pending U.S. Food and Drug Administration clearance -- could reduce reading time in interpreting DBT studies while maintaining or improving radiologist performance.

In a reader study involving 24 radiologists reading 260 DBT cases both with and without AI, the radiologists realized a 5.7% improvement in the area under the receiver operating characteristic curve, an 8% increase in sensitivity, a 6.9% increase in specificity, and a 7.2% reduction in recall rate with the use of AI. What's more, reading times were shortened by 52.7%.

"Therefore, reading 2D [mammography] plus DBT with AI improved the readers' performance while also reducing reading time compared to reading without AI," Conant told AuntMinnie.com.

Get all the details by attending this Monday morning presentation.