Lunit study shows higher HER2 diagnostic agreement between AI and pathologists

Lunit said Tuesday that it has published research showing that its AI biomarker platform, Lunit SCOPE, can improve HER2 (human epidermal growth factor receptor 2) diagnosis in advanced bile duct cancer. The study was published in Laboratory Investigation, the official journal of the United States and Canadian Academy of Pathology (USCAP) (Impact Factor 4.2).

Biliary tract cancer is a rare cancer with a poor prognosis. Recently, clinical research on HER2-targeted therapies, such as Enhertu (trastuzumab deruxtecan) and Ziihera (zanidatamab), has intensified, making HER2 testing increasingly important for selecting eligible patients for treatment.

Despite advances, HER2 immunohistochemistry (IHC) results can still vary depending on pathologists’ interpretation. This ongoing variability has led to persistent calls for a more consistent evaluation system.

Lunit published results of its HER2 diagnostic study for advanced biliary tract cancer using its AI biomarker platform, Lunit SCOPE, in the journal Laboratory Investigation (Impact Factor 4.2). (Courtesy of Lunit)
The study, conducted by researchers from Lunit, CHA University Bundang Medical Center, and CHA University Ilsan Medical Center, examined differences in HER2 IHC interpretation among pathologists for advanced bile duct cancer and assessed agreement between an AI model and pathologist consensus.

The team analyzed 309 HER2 IHC slides from 291 patients treated for advanced biliary tract cancer at CHA University Bundang Medical Center between 2019 and 2022. Three pathologists independently interpreted the slides using light microscopy and digital pathology. Results were then compared with Lunit SCOPE’s HER2 diagnosis.

The analysis revealed that the rate of identical readings among the three pathologists was 62.1 percent for optical microscopy and 63.4 percent for digital pathology, confirming inter-reader variability. In contrast, Lunit SCOPE showed higher concordance, matching the pathologists’ consensus results 83.5 percent of the time. Notably, the agreement rate between AI and pathologists was higher when using digital pathology than with optical microscopy.

Building on the findings of this study, Lunit plans to expand its research into more specialized areas, such as HER2-low expression. In addition, it aims to broaden the scope of application for its digital pathology-based AI solution through further multicenter collaborative research.

“HER2 diagnosis naturally varies among pathologists. This study demonstrates AI’s ability to minimize that variation, enhancing objectivity and reproducibility,” Lunit CEO Brandon Suh said. “Lunit is committed to setting the standard for HER2 diagnosis across multiple cancers, accelerating wider treatment access and driving meaningful advancements for patients.”

SOURCE: Korea Biomedical Review