Researchers have shown that AI support significantly improves radiologists' accuracy in detecting breast cancer during mammogram screenings, without increasing reading time.
Study: Influence of AI decision support on radiologists’ performance and visual search in screening mammography. Image Credit: Okrasiuk/Shutterstock.com
Published in PubMed, the study used eye-tracking technology to demonstrate how AI guidance helps radiologists focus more effectively on suspicious areas while maintaining their usual pace. By categorizing mammograms based on risk, the AI system enabled radiologists to allocate their attention more strategically, spending less time on clearly low-risk cases and scrutinizing higher-risk areas more closely.
Background
Mammography continues to play a vital role in early breast cancer detection, but even experienced radiologists can struggle to consistently spot subtle lesions. While previous research has shown AI can enhance diagnostic sensitivity, its effect on visual search behavior remains unclear.
To investigate, researchers at Radboud University Medical Center studied how AI decision support influences radiologists’ attention and accuracy. Using an eye-tracking system, they monitored 12 radiologists as they reviewed 150 mammograms—half cancer-positive, half negative—with and without AI input.
The AI categorized each case by risk level (low to very high) and flagged potentially suspicious regions. This setup allowed the researchers to analyze shifts in visual behavior and measure diagnostic performance. Their findings add to growing evidence that AI can support radiologists in making faster, more accurate distinctions between normal and abnormal findings.
AI Enhances Detection While Preserving Speed
The results were clear: radiologists caught more cancers when using AI assistance. Eye-tracking data showed that they spent more time examining actual lesions highlighted by the AI, while moving more quickly through low-risk scans. For instance, when the AI flagged a case as low-risk, radiologists appeared more confident and efficient. In contrast, high-risk scores prompted them to slow down and inspect more carefully.
Visual markers—circles for soft tissue abnormalities and diamonds for calcifications—acted as intuitive guides, drawing attention to areas needing closer review. This approach not only boosted accuracy but did so without extending the time spent per scan. In one case, a 67-year-old woman had a mass classified as “very high risk” by AI; radiologists were notably more likely to detect it when assisted by the system.
These findings show how AI can help radiologists prioritize high-yield regions, making their work both more effective and more efficient.
Balancing AI Support With Clinical Judgment
Despite the benefits, researchers warn against overdependence on AI. While rare, incorrect AI suggestions could still lead to missed diagnoses or unnecessary follow-ups. That’s why the team emphasizes the importance of high-accuracy AI models and proper training for radiologists to critically evaluate AI outputs.
Ongoing research aims to refine how and when AI input is delivered. For example, whether results should appear right away or only when requested. The team is also developing methods to signal when the AI’s confidence is lower, allowing for more cautious interpretation. These refinements could help maximize the benefits of AI while reducing potential downsides.
Conclusion
This study highlights AI’s ability to support radiologists by improving diagnostic accuracy and streamlining mammogram reviews. By guiding attention to higher-risk areas and speeding up review of low-risk cases, AI systems can help radiologists work more effectively without compromising quality.
Successful integration, however, depends on using trustworthy AI tools and ensuring radiologists are well-equipped to interpret their outputs. As screening volumes continue to rise globally, AI-powered support tools may become essential in helping healthcare systems meet growing demand while maintaining high standards of care.
Journal Reference
Gommers et al. (2025). Influence of AI decision support on radiologists’ performance and visual search in screening mammography. PubMed, 316(1), e243688. DOI:10.1148/radiol.243688. https://pubs.rsna.org/doi/10.1148/radiol.243688
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.