Retrospective comparison between single reading plus an artificial intelligence algorithm and two-view digital...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Traditional breast cancer screening relies on 2D full-field digital mammography (2D-FFDM) with double reading by radiologists.
• Artificial intelligence (AI) systems are increasingly being explored to enhance diagnostic accuracy in medical imaging.

Purpose of the Study:

• To evaluate the breast cancer detection rate using a single radiologist supported by an AI system.
• To compare this AI-assisted approach with the standard 2D-FFDM double-reading method.

Main Methods:

• A retrospective analysis of 161 biopsy-proven breast cancers using 3D tomosynthesis images.
• Comparison of AI algorithm performance against individual radiologists and consensus from double reading at different sensitivity thresholds.

Main Results:

• AI-assisted single reading showed significantly higher detection rates than single-reading 2D-FFDM.
• At higher sensitivity, AI was significantly more sensitive than individual radiologists and matched consensus sensitivity.
• Detection capability was consistent across different tumor types, grades, and breast densities.

Conclusions:

• Single reading with an AI algorithm demonstrates non-inferior sensitivity compared to standard 2D-FFDM double reading.
• AI support offers a viable alternative for enhancing breast cancer detection in tomosynthesis screening.