Workload reduction of digital breast tomosynthesis screening using artificial intelligence and synthetic mammography:...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Digital breast tomosynthesis (DBT) offers high sensitivity but requires extensive reading time.
• Synthetic mammography (SM) images, derived from DBT, are quicker to interpret and comparable to digital mammography (DM).
• Balancing screening sensitivity with radiologist workload is a critical challenge in breast cancer detection.

Purpose of the Study:

• To investigate the use of artificial intelligence (AI) for stratifying breast cancer screening examinations.
• To determine if AI can direct cases to either SM or DBT reading to minimize workload while maximizing accuracy.
• To compare the efficacy of AI-guided reading strategies against standard reading protocols.

Main Methods:

• Retrospective analysis of paired DM and one-view DBT from the Malmö Breast Tomosynthesis Screening Trial.
• Utilized the ScreenPoint Transpara 1.7 AI system to analyze DBT examinations.
• Simulated SM reading for low-risk cases (equivalent to DM) and used DBT reading for high-risk cases, exploring various single and double reading combinations.

Main Results:

• Double-reading DBT for the highest-risk 30% of cases and single-reading SM for the remainder detected 122 cancers.
• This approach matched the reading workload of standard DM double reading.
• Achieved 28% more cancer detections than DM double reading and captured 96% of cancers found with full DBT double reading.

Conclusions:

• AI can effectively stratify breast cancer screening examinations in a DBT program.
• High-risk cases benefit from DBT reading, while SM is sufficient for low-risk cases.
• This AI-driven strategy significantly increases cancer detection compared to DM alone, with only a minor increase in reading workload, warranting prospective studies.