Data from five studies involving Lunit’s SCOPE platform have demonstrated how artificial intelligence (AI)-based quantitative pathology can support research across different cancer types, further reinforcing the technology’s growing role in the field.
Results from the studies are being presented at the 2026 European Society for Medical Oncology (ESMO) congress, held in Madrid from 23-27 October.
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
According to Lunit, the studies involving SCOPE showcase the platform’s ability to reveal insights into themes such as biomarker differences, along with its role in assessing the relationship between tumour tissue characteristics and clinical outcomes. SCOPE is a digital pathology suite of multiple tools to help support immuno-oncology research.
In a large-scale Japanese study involving 993 patients with surgically resected non-small cell lung cancer (NSCLC), researchers compared PD-L1 tumour proportion scores (TPS) assessed independently by a pathologist and Lunit SCOPE PD-L1.
PD-L1 is a software module within Lunit’s broader SCOPE platform that applies AI-based digital pathology to analyse programmed death-ligand 1 (PD-L1 biomarker expression in cancer tissue samples.
The overall rate of agreement between a pathologist and Lunit’s platform was 73%, with the lowest concordance observed in tumours classified as PD-L1 negative by a pathologist.
Meanwhile, of 211 tumours with a TPS score <1% by the pathologist but TPS ≥1% by AI, 97% were classified as having low PD-L1 expression (TPS 1–10%) by AI. The findings highlight differences between manual and AI-based PD-L1 assessment, Lunit stated, particularly at low expression levels, while providing insights into areas that warrant further clinical investigation.
In another study involving 181 patients with pancreatic ductal adenocarcinoma (PDAC), researchers used Lunit SCOPE IO to quantify tumour-stroma ratios and endothelial cell proportions in pathology images and examined their associations with KRAS mutation status and survival outcomes.
Another software module within SCOPE, IO is designed to analyse Hematoxylin and Eosin (H&E)-stained whole slide images (WSIs) to evaluate the tumour microenvironment (TME), and other aspects including tumour-infiltrating lymphocytes (TILs), immune phenotypes, and spatial cellular patterns.
While higher tumour-stroma ratios were associated with worse progression-free survival overall in the PDAC study, the opposite association was observed in patients with KRAS G12D-mutated tumours. Higher endothelial cell proportions were also associated with favourable survival outcomes in tumours with lower stromal content, but not in those with higher stromal content.
According to Lunit, these findings highlight how the prognostic significance of tumour tissue characteristics can vary depending on molecular subtype and tissue context, suggesting the potential of AI-powered quantitative pathology to uncover “complex relationships” between tumour biology and patient outcomes.
Lunit’s CEO, Brandon Suh, said: “AI is helping researchers uncover new insights from pathology images across different cancer types.
“These studies reflect the diverse ways Lunit SCOPE can support cancer research. We will continue to support researchers and expand collaborations that advance our understanding of cancer biology and precision oncology.”
