Reducing Mammography Workload by Nearly 40 Percent? What a New Hybrid AI Study Reveals

Emerging research suggests that a hybrid radiologist/AI screening model — in which the level of certainty with AI findings determines the degree of radiologist involvement — may lead to a significant workload reduction without affecting breast cancer detection rates (CDRs).

In the study, recently published in Radiology, researchers assessed the hybrid radiologist/AI (Transpara, ScreenPoint Medical) approach for 41,469 mammography exams performed for 15,522 women (media age of 59). Recall decisions were made solely by the AI model when there was a degree of certainty with AI evaluation. In all other cases involving a lack of certainty with AI interpretation, there was double reading by radiologists, according to the study.

In comparison to the European standard of double reading by radiologists, the researchers found that use of the hybrid radiologist/AI screening approach resulted in a 38.1 percent reduction in mammography reading workload for radiologists with comparable CDRs (6.7 percent vs. 6.6 percent) and recall rates (23.9 percent vs. 23.6 percent).

“What we found is that with this hybrid reading strategy, radiologists only needed to read 61.9 percent of the cases, and the other ones were able to be read by AI alone (with) no human interference at all,” noted lead study author Sarah Verboom, MSc, Ph.D(c) in a recent interview with Diagnostic Imaging. “What we saw with this reading strategy that it had no effect on recall rate and cancer detection that stayed completely the same. So we were able to do workload reduction with no changes in the performance.”

During the interview, senior study author Ioannis Sechopoulos, Ph.D., said the next step will be examining the hybrid radiologist/AI screening model in a large prospective study involving mammography screening for 84,000 women in the Netherlands.

“This will give us the real evidence this approach would work, and we'll see, as Sarah said, what workload reduction and or performance we can get out of this hybrid mechanism. (This is) pretty exciting,” added Dr. Sechopoulos, the director of the Advanced X-Ray Tomographic Imaging Lab in the Department of Medical Imaging at the Radboud University Medical Center in Nijmegen, the Netherlands.

(Editor’s note: For related content, see “Reducing the Interval Breast Cancer Rate of Screening DBT: Can AI Have an Impact?,” “Expanded Breast Cancer Screening in Missouri Led to 45 Percent Higher Likelihood of Mammography Screening for Women on Medicaid” and “Mammography Study: AI Facilitates Greater Accuracy and Longer Fixation Time on Suspicious Areas.”)

For more insights from Dr. Sechopoulos and Ms. Verboom, watch the video below.