RSNA2022
27 November 2022
Authors
Gialias P., Brehl A K, Gustafsson H.
Artificial intelligence (AI) allows safe workload reduction in breast cancer screening: A retrospective study
Purpose
Retrospective simulation to test the potential of Transpara to safely reduce workload in the clinical workflow of a double-reading breast cancer screening program.
Materials and Method
Between September 2021 and February 2022, 15468 DM screening exams (including 53 screen-detected cancers) of 15468 women (age range: 40 to 74) were recorded in the biennial breast cancer screening program in Östergötland county in Sweden. Each DM exam was double read independently by two breast radiologists and processed by Transpara (version 1.7.1, ScreenPoint Medical). Transpara assigns a score on a scale of 1-10 to each exam with increasing likelihood of cancer. In a retrospective simulation, Transpara was implemented as a triaging tool; exams with a low risk (Transpara 1-7) were selected for single reading, while exams with an elevated risk (Transpara score 8 or higher) were selected for double-reading according to the standard clinical protocol.
Results
Implementing the triaging strategy with a threshold of Transpara score 7 for single vs. double reading, 10473 exams (67.7%) received a score between 1 and 7. The triaging strategy would lead to 33.8% workload reduction. Overall, 52 out of 53 screen-detected cancers were picked up by Transpara and received a score higher than 7. The one cancer that was missed by Transpara and received a score of 4 was detected by single reading.
Conclusion
Replacing one reader in a breast cancer screening workflow with Transpara for the low risk cases could safely reduce the workload by 33.8% with no cancers being missed.