Lunit, a medical artificial intelligence (AI) company, announced on the 20th that it will present five research abstracts using its AI biomarker platform Lunit SCOPE at the 2026 European Society for Medical Oncology (ESMO) Congress, which will take place in Madrid, Spain, from the 23rd to the 27th.
The main presentations will feature findings from AIforHER, a global AI project focused on HER2 analysis. Led by French cancer center ICO, the project brings together global pharmaceutical companies and medical institutions, including AstraZeneca and Daiichi Sankyo, to develop and validate AI algorithms for personalized breast cancer treatment.
HER2 is a protein involved in cancer cell growth and proliferation, and breast cancer treatment options vary depending on its expression level. HER2-ultralow refers to cases in which the protein is present at extremely low levels. The importance of accurate assessment has grown as some antibody-drug conjugates (ADCs) have recently become treatment options for patients in this category.
Lunit will present findings from Study A, the first study to compare AI solutions from multiple companies for HER2-ultralow breast cancer. The company is also participating in the follow-up Study B, which currently includes only two of the participating companies, to evaluate AI’s potential to identify eligible patients in real-world clinical settings.
Lunit will also disclose findings from a study on non-small cell lung cancer conducted with Japan’s National Cancer Center. When pathologists and AI independently evaluated tissue samples from 993 patients, the overall agreement rate for PD-L1 expression levels was 73%. Agreement reached 80% for expression levels of 50% or higher and 95% for levels of 1% to 49%, but fell to 51% for levels below 1%. The researchers said further investigation is needed to determine how these differences in assessment affect actual treatment outcomes.
Another study will examine how tissue surrounding tumors and genetic mutations affect prognosis in 181 patients with pancreatic ductal adenocarcinoma.
The analysis found that patients with a higher stromal proportion had relatively shorter progression-free survival. Median overall survival varied by KRAS mutation status: 43.8 months for patients without mutations, 19.5 months for those with other mutations and 15.1 months for those with the G12D mutation. However, among patients with the G12D mutation, those with more stromal tissue had longer progression-free survival.
“These studies are meaningful because they examine the role AI can play in areas such as HER2-ultralow expression, where assessment through microscopy alone can be challenging,” said Lunit CEO Seo Beom-seok. “We will expand global collaborations to more precisely identify patients who are eligible for cancer treatment.”
