12 Jul Mount Sinai Scientists Develop Test for Lung Cancer Biomarker Detection from Common Pathology Slides
MedicalResearch.com Interview with:
Dr. Campanella
Gabriele Campanella, PhD
Assistant Professor
Windreich Department of Artificial Intelligence and Human Health
Icahn School of Medicine at Mount Sinai
MedicalResearch.com: What is the background for this study?
Response: Lung cancer is the most lethal cancer in the US. Lung adenocarcinoma (LUAD) is the most common form of lung cancer with an incidence of over 100k per year in the US. EGFR mutations are common driver mutations in LUAD, and importantly, these mutations can be targeted by TKI therapy, which has high response rates. Because of this, EGFR testing via NGS (Next Generation Sequencing) is considered mandatory by guidelines for any LUAD diagnosis.
In high-resource settings, rapid EGFR testing is done while waiting for confirmation via NGS. This is because NGS takes about 2 weeks on average, while the rapid testing has a median TAT of 2 days. Early treatment decisions could be made based on the rapid test results. Rapid tests have some important drawbacks, most notably, it exhausts tissue. In lung cancer, tissue is scarce in the first place, and up to 25% of cases, after rapid testing there is not enough tissue for NGS. In those circumstances, patients have to be biopsied again, which adds unnecessary risk for the patient. Even worse, in some cases, the NGS is never done. A non-tissue-exhaustive computational biomarker could be used instead of the tissue-based rapid test.
MedicalResearch.com: What are the main findings?
Response: In this study we tested the idea of using an H&E digital slide-based computational biomarker to replace the rapid testing in certain conditions. Throughout internal validation and external multi-center validation we found that it is possible to predict EGFR mutations from pathology slides with a performance that allows for clinical application. We further tested real world deployment in a silent trial and found that 43% of rapid tests could be avoided while keeping the gold standard performance of the current clinical workflow.
MedicalResearch.com: Could the EGFR mutations be detected from regularly H&E slides or was special processing, ie frozen sections required?
Response: Only standard H&E slides were considered.
MedicalResearch.com: What should readers take away from your report?
Response: Our work shows that H&E-based computational biomarkers can be deployed clinically with a benefit for the patients and a more efficient clinical workflow. Importantly, we thoroughly tested the model internally, and across multiple institutions, to ensure the generalizability and stability.
MedicalResearch.com: What recommendations do you have for future research as a results of this study?
Response: From the methodological side, for the first time we propose end-to-end task-specific fine tuning of a foundation model, showing superior performance than traditional methods relying on frozen foundation models.
In addition, we perform the first silent trial for a computational pathology AI model. We believe this is a necessary step in proving the safety and effectiveness of these models and enable potential regulatory approvals.
MedicalResearch.com: Is there anything else you would like to add? Any disclosures?
Response: In the future, we plan to expand our analysis to additional computational biomarkers across cancers. In parallel, we will study the benefits of deploying computational biomarkers in low-resource settings where NGS access is limited or absent.
Citation: Campanella, G., Kumar, N., Nanda, S. et al. Real-world deployment of a fine-tuned pathology foundation model for lung cancer biomarker detection. Nat Med (2025). https://doi.org/10.1038/s41591-025-03780-x
https://www.nature.com/articles/s41591-025-03780-x
===
The information on MedicalResearch.com is provided for educational purposes only, and is in no way intended to diagnose, cure, or treat any medical or other condition. Some links are sponsored. Products and claims are not warranted or endorsed.
Always seek the advice of your physician or other qualified health and ask your doctor any questions you may have regarding a medical condition. In addition to all other limitations and disclaimers in this agreement
Last Updated on July 12, 2025 by Marie Benz MD FAAD
