Artificial Intelligence for Reducing Workload in Breast Cancer Screening with Digital Breast Tomosynthesis.

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence

Background:

• Digital breast tomosynthesis (DBT) offers higher accuracy than digital mammography but requires more interpretation time.
• Artificial intelligence (AI) presents a potential solution to enhance the efficiency of reading screening examinations.
• Optimizing radiologist workflow is crucial for managing increasing screening volumes.

Purpose of the Study:

• To assess the efficacy of an AI model in reducing the workload associated with interpreting digital breast tomosynthesis screening examinations.
• To determine if AI can accurately identify and filter out normal DBT examinations, allowing radiologists to focus on potentially abnormal cases.

Main Methods:

• A retrospective analysis of 13,306 DBT examinations from 9,919 women was conducted.
• An AI model was trained, validated, and tested to classify cancer-free examinations for potential dismissal from the screening worklist.
• The AI system's performance was further evaluated in a reader study involving five breast radiologists.

Main Results:

• The AI model simulated a 39.6% reduction in radiologist workload.
• Noninferior sensitivity (90.0% vs 90.8%) and a 25% lower recall rate (6.9% vs 9.2%) were observed with AI assistance.
• The standalone AI model demonstrated a higher area under the receiver operating characteristic curve (AUC) compared to the average radiologist (0.84 vs 0.81).

Conclusions:

• AI models can effectively identify normal digital breast tomosynthesis screening examinations.
• Implementing AI in clinical workflows can significantly decrease the number of examinations requiring radiologist interpretation.
• AI holds promise for improving the efficiency and potentially the accuracy of breast cancer screening interpretation.