Lunit, a leader in AI-powered cancer diagnostics, is set to make a significant impact at the European Congress of Radiology (ECR) 2025, held in Vienna, Austria, from February 26 to March 2. The company will present 15 research abstracts, with an impressive 13 selected for oral presentations, marking its largest and most influential presence at the conference to date.
The high number of accepted presentations underscores Lunit’s growing leadership in AI-driven radiology. The research findings highlight the potential of AI to revolutionize breast cancer detection, optimize radiology workflows, and ultimately improve patient outcomes. Crucially, the studies demonstrate AI’s capability to perform several key tasks, including:
- Replacing one radiologist in double reading of mammograms.
- Identifying subclinical breast cancer years before a clinical diagnosis.
- Maintaining strong detection performance even with varying image quality.
Key Studies and Findings
Two key studies presented at ECR 2025 provide particularly compelling findings about Lunit INSIGHT MMG’s clinical utility in mammography.
Replacing a Radiologist with AI for Independent Double Reading
One study, led by Marie Burns Bergan from BreastScreen Norway, evaluated the impact of using Lunit INSIGHT MMG to replace one radiologist in a large-scale mammography screening program. The research analyzed over one million mammography screenings, testing various AI thresholds to determine the system’s role in maintaining detection accuracy while minimizing radiologists’ workload. Some key findings:
- AI successfully detected up to 79.9% of screen-detected cancers at a 10% positivity threshold.
- At a lower, 5% threshold, AI maintained 75.5% cancer detection while also identifying 5.7% of interval cancers, which often escape traditional screening methods.
- AI integration reduced radiologists’ reading workload by 50%, thereby optimizing resources without compromising accuracy.
This study supports the idea that Lunit INSIGHT MMG can serve as a reliable second reader, supporting improved efficiency and addressing radiologist shortages in high-volume screening programs.
Detecting Subclinical Breast Cancer with AI
A second study, spearheaded by Jonas Gjesvik from BreastScreen Norway, investigated Lunit INSIGHT MMG’s ability to identify subtle indicators of cancer risk years ahead of a formal diagnosis. This analysis involved 116,000 women participating in a long-term mammography screening program. The study showed that:
- AI assigned higher risk scores to breasts that subsequently developed cancer, even up to six years before a diagnosis.
- AI scores for screen-detected breast cancers increased dramatically from 19.2 to 82.7 across successive screenings, reflecting early disease detection.
- For interval cancers, AI identified elevated risk earlier than traditional methods, paving the way for more personalized screening strategies.
The results underscore the promise of using AI to enhance breast cancer risk prediction, ultimately shifting towards a personalized, risk-based screening model.
Broader Impact and Future Directions
In addition to the two highlighted studies, Lunit will present an additional 13 research abstracts. These further demonstrate the wide-ranging applications of AI in radiology, with focus on enhancing diagnostic accuracy, optimizing radiology workflows, and supporting radiologists worldwide. “This year’s ECR is a landmark moment for Lunit, with our highest-ever number of oral presentations, underscoring our leadership in AI-powered radiology research,” said Brandon Suh, CEO of Lunit. “From replacing a radiologist in double-reading settings to predicting breast cancer years in advance, these findings reinforce AI’s potential to transform breast cancer detection, optimize radiology workflows, and ultimately improve patient outcomes.”
