Lunit to unveil 12 AI-powered oncology studies at ASCO 2025

Lunit, a Korea medical AI imaging company, said it will present 12 research abstracts at the upcoming 2025 American Society of Clinical Oncology (ASCO) Annual Meeting, taking place from May 30 to June 3 in Chicago.

The company will showcase the expanding clinical utility of its AI biomarker platform Lunit SCOPE, with key findings spanning HER2-positive biliary tract cancer, lung cancer, and gastric cancer.

Notably, the company is set to present the results of a study conducted between Lunit and Japan’s National Cancer Center East (NCCE), evaluating the efficacy of Enhertu (ingredient: trastuzumab deruxtecan) in HER2-positive biliary tract cancer patients.

The study used Lunit SCOPE uIHC to analyze immunohistochemistry (IHC) slides from 288 patients. The AI assessed both HER2 expression intensity and its cellular localization, particularly distinguishing between membrane and cytoplasmic expression patterns.

The results revealed that membrane-specific HER2 expression, as detected by Lunit’s AI, was more predictive of treatment outcomes than expression intensity alone.

Among 29 patients treated with Enhertu, those with high membrane-specific HER2 expression had a median progression-free survival (mPFS) of 11 months, compared to just 4.2 months in those with lower membrane specificity.

Notably, the AI identified additional treatment-responsive patients who would have been missed by traditional HER2 assessment methods, suggesting enhanced patient stratification via AI-guided analysis.

In a second joint study with NCCE, Lunit evaluated its PD-L1 scoring AI, Lunit SCOPE PD-L1, in a large-scale, prospective observational study involving 847 non-small cell lung cancer (NSCLC) patients and 102 small cell lung cancer (SCLC) patients.

The AI’s PD-L1 scoring results were compared against assessments from three expert pathologists.

The study reported a 70 percent concordance rate between AI and human pathologists, reinforcing the AI’s robustness across a large, independent patient population.

While AI-based PD-L1 analysis has been previously studied, this marks one of the largest head-to-head validations of its performance in real-world clinical settings.

Lunit will also present a study focusing on claudin 18.2 (CLDN18.2) expression and immune phenotype prediction in gastric cancer.

As CLDN18.2-targeting therapies gain traction as a new treatment avenue, widespread implementation has been limited by sample constraints, high costs, and time-consuming diagnostics.

To address this, Lunit developed an AI model capable of predicting CLDN18.2 expression from standard hematoxylin and eosin (H&E) slides alone. The model demonstrated a high predictive performance, with an area under the receiver operating characteristic curve (AUROC) of 0.751.

CLDN18.2 expression was also found to correlate with immune phenotype.

The study highlighted that CLDN18.2-negative patients with an “inflamed” immune phenotype benefited significantly from combination immunotherapy and chemotherapy, showing marked improvements in both progression-free survival (PFS) and overall survival (OS) compared to chemotherapy alone.

“Our ASCO 2025 presentations demonstrate not only the technological sophistication of Lunit’s AI but also its tangible contributions to clinical oncology across multiple cancer types,” Lunit CEO Brandon Suh said. “We are committed to continuously advancing the Lunit SCOPE platform to meet unmet medical needs and deliver precision medicine tailored to each patient.”