Medical artificial intelligence (AI) company Lunit announced on May 23 that it will present 12 research findings utilizing its AI biomarker platform, Lunit Scope, at the 2025 American Society of Clinical Oncology (ASCO 2025), which will be held in Chicago, United States, from May 30 to June 3.
The most significant study Lunit will present at ASCO 2025 is a collaborative research project with the National Cancer Center East (NCCE) of Japan, which predicts the therapeutic efficacy of Enhertu (ingredient: trastuzumab deruxtecan) in patients with HER2 (human epidermal growth factor receptor 2) positive biliary tract cancer.
In this study, Lunit Scope uIHC was applied to analyze immunohistochemistry (IHC) slide images from 288 patients with HER2-positive biliary tract cancer. The platform quantitatively evaluated the distribution of HER2 expression by both the intensity and the subcellular localization (such as cell membrane and cytoplasm) of the cancer cells.
The results showed that Lunit AI could better predict patient response to treatment not only by the intensity of HER2 expression but also by the specificity of HER2 expression on the cell membrane. Among 29 patients treated with Enhertu, those with high cell membrane specificity had a median progression-free survival (mPFS) of 11.0 months, compared to 4.2 months for those with low cell membrane specificity, indicating improved treatment outcomes.
Notably, among the patients identified by AI, there were additional patients who would not have been detected using conventional HER2 expression intensity measurement methods. This suggests that AI-based analysis of cell membrane specificity can identify a greater number of patients who may respond to Enhertu treatment than traditional methods.
The second collaborative study between Lunit and NCCE compared the results of Lunit Scope PD-L1, an AI-based tool, with the assessments of three pathologists for 847 non-small cell lung cancer (NSCLC) patients and 102 small cell lung cancer (SCLC) patients enrolled in a prospective observational study.
The overall concordance rate between the AI and the pathologists was found to be 70%. While AI analysis of PD-L1 has been reported continuously, the significance of this study lies in validating the performance of AI in a large, independent patient cohort.
Finally, Lunit will present research findings on the prediction of 'Claudin18.2 (CLDN18.2)' protein expression and immune phenotype in gastric cancer patients. While therapies targeting Claudin18.2 have emerged as a new alternative for gastric cancer treatment, practical limitations such as insufficient samples, additional costs, and the time required for testing have persisted.
To address these challenges, Lunit developed an AI model that predicts Claudin18.2 expression using only basic H&E slides. In the gastric cancer patient cohort, the model achieved a high AUROC value of 0.751 for predicting Claudin18.2 positivity.
This result demonstrates the potential to pre-select Claudin18.2-positive patients using only H&E slides in the future.
Furthermore, Claudin18.2 expression status is associated with response to immune checkpoint inhibitor therapy. In the patient group found to be Claudin18.2-negative but with an inflamed immune phenotype, treatment with combined immuno-chemotherapy resulted in significantly longer progression-free survival (PFS) and overall survival (OS) compared to chemotherapy alone.
This indicates that AI-predicted Claudin18.2 expression and immune phenotype information could serve as valuable tools to assist treatment decisions for gastric cancer patients.
Seo Beomseok, CEO of Lunit, stated, "At ASCO 2025, Lunit has proven both the technological capabilities of its AI and its potential clinical contributions across various cancer types. We will continue to advance the Lunit Scope solution to address unmet needs in cancer treatment and provide optimized therapies for individual patients."
Meanwhile, Lunit has participated in ASCO, one of the world's three largest cancer conferences, every year since 2019 to present the latest research findings related to biomarkers. This year, the company plans to present a record 12 research abstracts, the highest number among medical AI companies worldwide.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
