Combining DBT with AI-based support may be key to cost-effective breast screening programs

Radiologists’ reading performance significantly improves when using an artificial intelligence-based support system during digital breast tomosynthesis screening exams, according to research published Tuesday.

Institutions using DBT-based programs typically combine it with digital or synthetic mammography with two views per breast, authors explained in Radiology. This approach, however, produces more images for rads’ to read and increases patients’ radiation dose. Some experts have reported success using a single view to overcome these issues.

This proved true for German and Dutch researchers. They found breast subspecialists’ interpretations significantly improved when using AI support to manage these single-view exams.

Although there was no change in specificity or reading times, they noted, CDS is a good option for organizations looking to implement DBT screening programs. 

“Overall, our study adds to the growing body of evidence on the improvement of screening outcomes by adding an AI system to DBT for decision and navigation support, suggesting that use of single-view DBT combined with AI is a feasible DBT screening implementation option,” Marta C. Pinto, of Radboud University Medical Center’s Department of Medical Imaging, and co-authors wrote July 6.

The retrospective observer study utilized single-view bilateral mediolateral oblique exams and corresponding synthetic 2D images gathered over a nearly two-year period. Fourteen breast specialists interpreted 190 DBT exams across two sessions with AI decision support and without assistance.

Upon comparison, radiologists’ exam-based average area under the receiver operating characteristic curve score was higher with AI support than without it (0.88 versus 0.85). Rads’ average sensitivity also jumped with technological assistance (86% vs. 81%).

“Using this DBT and AI setup could allow for a more cost-effective screening program with higher performance and a lower workload for the radiologists,” Pinto et al. added.

Two University of Michigan radiology experts responded to these results in an editorial published alongside the study. Heang-Ping Chan, PhD, and Mark A. Helvie, MD, both acknowledge single-view DBT protocols would reduce radiation dose by about half while also cutting reading times.

But they noted current data don’t yet support using this strategy over conventional breast screening approaches.

“There is no strong evidence to date to support a single-view DBT with similar performance to a two-view protocol that includes an orthogonal DBT or DM view,” the pair wrote. “Large-scale studies in the screening population with direct comparison to the current two-view protocols are needed to investigate whether this single-view DBT with AI CAD protocol can match or surpass the two-view protocols with or without AI CAD.

Read Pinto and colleagues’ full study here.

Around the web

The consensus guidelines are associated with a high specificity, new research confirmed. 

Researchers think their work could play an important role in the future of imaging technology. 

The weeklong course will include live lectures, group discussions and more.

Trimed Popup