Assessment of Interval Breast Cancer Screening Mammograms Using Radiologists and AI

Overview

This retrospective study evaluated whether an AI system can identify breast cancer signs in interval breast cancer (IBC) screening mammograms and compared its performance with experienced radiologists. Among 119 interpretable IBC mammograms, the AI system demonstrated potential to detect signs of cancer that were missed during routine screening, highlighting its promise to reduce IBC incidence.

Background

Mammography screening programs (MSPs) reduce breast cancer mortality by enabling early detection, but interval breast cancers—diagnosed between scheduled screenings—remain a challenge due to factors like dense breast tissue and radiologist interpretation errors. IBCs tend to have more aggressive features and worse outcomes, making their reduction a key quality indicator of MSPs. Artificial intelligence (AI) offers a promising approach to enhance detection sensitivity without increasing workload, but evidence on AI’s ability to detect IBCs is limited, especially in Switzerland.

Data Highlights

Key Findings

• The AI system ProFound AI® analyzed 119 IBC mammograms compatible with its platform.
• Interval breast cancers were defined as invasive or in situ breast cancers diagnosed within 24 months after a normal screening mammogram.
• Three experienced radiologists independently reviewed the IBC mammograms retrospectively without AI assistance.
• Older mammography devices (pre-2014) produced images incompatible with the AI system, limiting AI analysis to a subset of cases.
• The AI system provides a case score (0-100) indicating likelihood of breast cancer and a risk score categorizing future cancer risk.

Clinical Implications

Integrating AI into mammography screening workflows may enhance early detection of interval breast cancers, potentially reducing their incidence and improving patient outcomes. However, compatibility of imaging equipment with AI systems is essential to maximize benefit. Radiologists’ expertise remains critical, but AI can serve as an adjunct tool to identify subtle signs missed during routine screening.

Conclusion

This study supports the potential of AI to detect breast cancer signs in interval screening mammograms, complementing radiologist review and contributing to improved screening program effectiveness. Further prospective studies are warranted to validate these findings and optimize AI integration.

References

1. Niraula et al 2021 -- Interval Breast Cancer and Screening Effectiveness
2. European Guidelines 2020 -- Breast Cancer Screening Quality Indicators
3. iCAD ProFound AI System Documentation 2023
4. Cancer Registry of Eastern Switzerland 2010-2019 Data