Non-small-cell lung cancer (NSCLC) is a leading cause of death and is often diagnosed at advanced stages1. Current guidelines recommend molecular testing to identify driver mutations for targeted therapies2. This study aims to develop a rapid diagnostic workflow combining instant digital microscopy and artificial intelligence (AI) to predict the most likely driver mutation based on tissue morphology and clinical data. Samples were collected via EBUS/TBNA and thoracentesis from lung cancer patients. Fresh samples were analyzed with Vivascope, a confocal laser microscope able to provide native digital tissue images, without damage of biological sample. Pathologists ensured tissue adequacy and availability for molecular testing and manually annotated images. Images were used to train an AI model to predict the most likely driver mutation. Predicted KRAS, EGFR and BRAF mutations were rapidly tested with Idylla (Biocartis); others underwent NGS. AI predictions were compared with molecular results obtained from paired paraffin sections. The AI model demonstrated a good performance in recognizing KRAS, EGFR, BRAF mutations based on digital images. Compared with the ground truth, the algorithm predicted driver mutations in NSCLC samples with 71,5% of accuracy guiding the selection of the most appropriate molecular test (PCR or NGS). Validation on a broader, independent cohort is currently ongoing. This study firstly demonstrated the feasibility and value of combining digital microscopy and AI in assisting pathologists in NSCLC diagnosis and molecular test selection. The rapid workflow enables faster clinical decisions and sustains accessibility to targeted treatment even in centers with limited technological resources. It could be integrated into clinical practice to enhance diagnostic accuracy and patient outcomes. 1. Sung H et al Cancer J Clin. 2021;71(3):209-249. doi:10.3322/caac.21660 2. Gregory K, et al. NCCN Guidelines Version 3.2025 Non-Small Cell Lung Cancer. 2025
AI-Supported Digital Pathology for Lung Cancer: A Rapid Workflow to Guide Molecular Profiling / Mercurio, Anastasia; Morano, Vittoria; Maricchiolo, Giulia; Frasca, Luca; Davoli, Daniele; Maria Di Matteo, Francesco; Crescenzi, Anna. - (2025). ( X Congresso Triennale SIAPeC-IAP. L'Anatomia Patologica nell'Era Digitale: Medicina di Precisione, Big Data e Intelligenza Artificiale Milan, Italy ).
AI-Supported Digital Pathology for Lung Cancer: A Rapid Workflow to Guide Molecular Profiling
Anastasia Mercurio;Vittoria Morano;Giulia Maricchiolo;Luca Frasca;Anna Crescenzi
2025
Abstract
Non-small-cell lung cancer (NSCLC) is a leading cause of death and is often diagnosed at advanced stages1. Current guidelines recommend molecular testing to identify driver mutations for targeted therapies2. This study aims to develop a rapid diagnostic workflow combining instant digital microscopy and artificial intelligence (AI) to predict the most likely driver mutation based on tissue morphology and clinical data. Samples were collected via EBUS/TBNA and thoracentesis from lung cancer patients. Fresh samples were analyzed with Vivascope, a confocal laser microscope able to provide native digital tissue images, without damage of biological sample. Pathologists ensured tissue adequacy and availability for molecular testing and manually annotated images. Images were used to train an AI model to predict the most likely driver mutation. Predicted KRAS, EGFR and BRAF mutations were rapidly tested with Idylla (Biocartis); others underwent NGS. AI predictions were compared with molecular results obtained from paired paraffin sections. The AI model demonstrated a good performance in recognizing KRAS, EGFR, BRAF mutations based on digital images. Compared with the ground truth, the algorithm predicted driver mutations in NSCLC samples with 71,5% of accuracy guiding the selection of the most appropriate molecular test (PCR or NGS). Validation on a broader, independent cohort is currently ongoing. This study firstly demonstrated the feasibility and value of combining digital microscopy and AI in assisting pathologists in NSCLC diagnosis and molecular test selection. The rapid workflow enables faster clinical decisions and sustains accessibility to targeted treatment even in centers with limited technological resources. It could be integrated into clinical practice to enhance diagnostic accuracy and patient outcomes. 1. Sung H et al Cancer J Clin. 2021;71(3):209-249. doi:10.3322/caac.21660 2. Gregory K, et al. NCCN Guidelines Version 3.2025 Non-Small Cell Lung Cancer. 2025I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
