Over-detection and over-surveillance in breast screening: current status and the potential for artificial...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Breast cancer screening saves lives but leads to over-detection of benign conditions, causing patient anxiety and increasing healthcare costs.
• Current screening methods like mammography and MRI often have reduced specificity, leading to false positives and unnecessary biopsies.

Purpose of the Study:

• To review breast imaging technologies, the impact of over-detection, and the role of AI in optimizing screening.
• To evaluate how AI can improve lesion stratification and personalize follow-up strategies in breast cancer screening.

Main Methods:

• Synthesis of evidence on screening technologies (mammography, DBT, ultrasound, MRI) and their limitations.
• Analysis of the multidimensional impact of over-detection, including psychosocial and economic burdens.
• Evaluation of AI's capacity to enhance sensitivity, specificity, and risk-adapted workflows.

Main Results:

• AI-assisted reading maintains or improves cancer detection while lowering recall rates and workload.
• AI enables finer risk stratification, particularly for BI-RADS 4 lesions, reducing unnecessary interventions.
• Refining Breast Imaging Reporting and Data System (BI-RADS) thresholds could decrease unnecessary procedures without compromising cancer detection.

Conclusions:

• Integrating validated AI into breast cancer screening can maximize benefits and minimize harm.
• A patient-centered strategy using AI can optimize follow-up intervals and support risk-adapted workflows.
• AI advancements can lead to more efficient and personalized breast cancer screening.