Lunit, a South Korean AI healthcare company, in collaboration with pharmaceutical giant AstraZeneca, will present groundbreaking research on an artificial intelligence model that predicts EGFR mutations in non-small cell lung cancer (NSCLC) patients at the upcoming American Association for Cancer Research (AACR) Annual Meeting 2025 in Chicago.
The study showcases the development and validation of the Lunit SCOPE Genotype Predictor, an AI-powered deep learning model that can predict Epidermal Growth Factor Receptor (EGFR) mutations directly from standard hematoxylin and eosin (H&E)-stained tissue samples, potentially transforming the diagnostic pathway for NSCLC patients.
Addressing Critical Gaps in Molecular Testing
EGFR mutation testing is essential for determining optimal treatment strategies for NSCLC patients, particularly for selecting appropriate targeted therapies. Despite clinical guideline recommendations, many patients worldwide remain untested due to logistical challenges, resource limitations, and access barriers.
Current molecular testing methods often require specialized equipment, expertise, and substantial tissue samples, creating bottlenecks in the diagnostic process. The Lunit-AstraZeneca collaboration aims to address these challenges by developing an AI solution that can rapidly predict mutation status from routine pathology slides that are already part of standard diagnostic workflows.
Unprecedented Scale and Diversity in AI Development
What distinguishes this research is the unprecedented scale and diversity of the training dataset. The study utilized more than 12,000 pathology slides from NSCLC patients across multiple countries, including the United States, China, and South Korea. The dataset included over 4,500 EGFR-mutated samples and more than 7,500 wild-type samples, representing the largest and most diverse training cohort for this application to date.
"Previous AI models designed to predict mutations from pathology images have faced limitations in real-world clinical applications due to limited training data and lack of validation," explained Brandon Suh, CEO of Lunit. "By leveraging Lunit AI, we have demonstrated that routine pathology slides can serve as a powerful tool to predict EGFR mutations with high accuracy."
Robust Performance Across Clinical Variables
A key strength of the Lunit SCOPE Genotype Predictor is its consistent performance across various clinical variables that typically challenge AI systems in healthcare. The model maintained reliable prediction capabilities regardless of specimen types (surgical resections or biopsies), specific EGFR mutation subtypes, different slide scanners, and varying scan magnifications.
This robust performance across diverse clinical settings suggests strong potential for real-world deployment in various healthcare environments globally, including resource-limited settings where molecular testing infrastructure may be lacking.
Clinical Implications for Precision Oncology
The AI tool is designed to rapidly and cost-effectively predict NSCLC driver mutations from H&E-stained tissue samples, potentially enabling clinicians to prioritize patients for confirmatory molecular testing and expedite treatment decisions.
"This study is a testament to the real-world potential of AI in precision oncology," Suh noted. "This could help clinicians prioritize molecular testing for NSCLC patients, ensuring that patients receive targeted therapy without unnecessary delays."
For patients with EGFR mutations, targeted therapies such as EGFR tyrosine kinase inhibitors (TKIs) have demonstrated significant improvements in progression-free survival compared to conventional chemotherapy. Faster identification of these patients could therefore lead to more timely initiation of appropriate targeted treatments.
Future Directions and Implementation
While the AI model shows promise as a screening tool, the researchers emphasize that it is intended to complement rather than replace conventional molecular testing. The technology could be particularly valuable in settings where molecular testing resources are limited or where rapid preliminary results could guide clinical decision-making while awaiting confirmatory testing.
The presentation at AACR 2025 represents a significant milestone in the ongoing collaboration between Lunit and AstraZeneca in the field of AI-driven precision oncology. The companies continue to explore additional applications of AI in cancer diagnostics and treatment selection.
Healthcare professionals and researchers interested in learning more about this study can visit Lunit at AACR 2025 Booth #2843 from April 25 to 30 in Chicago, Illinois.
