Faster treatment for lung cancer with AI
A new AI method can detect EGFR mutations in lung cancer at an early stage, helping speed up access to the right treatment. Researchers at the University of Gothenburg contributed to the work.
The results of the international study, led by Memorial Sloan Kettering Cancer Center in New York, have been published in Nature Medicine.
“Through this study, we’re helping make cancer diagnostics faster and more accessible. We show that AI can identify EGFR mutations with very high accuracy, directly from digital images of tissue sections,” says Levent Akyürek, professor of pathology at the Sahlgrenska Academy, University of Gothenburg, and co-author of the study.
Faster results
EGFR mutations are common in non-small cell lung cancer and are key to determining which treatment a patient should receive. Today’s molecular testing is complex, often takes several weeks, and requires large amounts of tumor tissue.
The AI model, called EAGLE, can instead analyze standard histology images (H&E) and accurately predict the presence of EGFR mutations. This could reduce the need for molecular testing.
The model was tested alongside routine clinical care without influencing it. Results showed that AI-based screening could reduce the need for rapid molecular tests by up to 43 percent — without compromising patient safety.
Saving tissue – and time
In many cases, the small tissue samples collected during biopsy aren’t enough for all required tests. AI could help identify which samples truly need deeper analysis, preserving tissue and reducing the need for additional biopsies.
“This technology has strong potential — especially in healthcare systems where access to molecular testing is limited. It can help make modern diagnostics more equitable,” says Akyürek. “In Sweden, all cases of non-small cell lung cancer are automatically tested. But in the US, testing often depends on a patient’s insurance coverage.”
The EAGLE model was trained on more than 8,000 tumor images from across several countries. Sahlgrenska University Hospital contributed digital pathology images, and researchers from Chalmers University of Technology and Sahlgrenska were also involved.
Because the study shows that the method works in real clinical settings, the model may be eligible for clinical approval. The researchers behind EAGLE now plan to develop similar AI tools to detect other genetic changes that influence cancer treatment decisions.
