• Lunit SCOPE IO, an AI therapeutic biomarker platform, can analyze immune phenotype of NSCLC patients at a higher accuracy level compared to pathologists

• With pathological consensus, AI-based analysis can be utilized as a novel biomarker of immunotherapy in the digital pathology era

• Findings to be presented at USCAP online poster presentation, Monday, March 15

Lunit today announced an upcoming abstract presentation featuring its AI-based predictive analysis of immune phenotype at the 2021 United States and Canadian Academy of Pathology (USCAP) Annual Meeting, to be held online on March 13-18.

Lunit, a leading medical AI startup, focuses on developing AI biomarkers that accurately predict cancer patients’ response to immunotherapy, based on digital analysis of histopathological whole slide images. With its proprietary deep learning algorithm, Lunit is developing an AI therapeutic biomarker platform called ‘Lunit SCOPE IO’.

In the abstract, Lunit presents the performance of Lunit SCOPE IO in the analysis of whole-slide images of non-small cell lung cancer (NSCLC) patients. The abstract, titled, ‘Development and validation of deep learning-based pathologic classification of immune phenotype in non-small cell lung cancer,’ will be presented during the poster session on Monday, March 15, 2:15 p.m. Pacific Time.

According to recent studies, Lunit SCOPE IO showed remarkable performance in classifying the immune phenotypes of NSCLC patients, by analyzing their H&E whole slide images. In the upcoming USCAP abstract presentation, the researchers share the findings from further validation of the AI’s performance, by directly comparing the AI detection results with that of pathologists’ diagnostic results.

By analyzing H&E whole slide images of NSCLC patients, the AI software accurately predicted three immune phenotypes (3-IP) by 92.4%. This showed a higher rate compared to the 3-IP annotation made by independent pathologists with the same validation set, which showed an 80.5% average concordance rate.

“Development of consensus and utilizing AI-based analysis to classify immune phenotypes are of high interest due to the lack of pathologic consensus for the classification of tumor infiltrating lymphocyte(TIL) distribution, which is the key factor in prediction of response to immunotherapy,” said Dr. Chan-Young Ock, Chief Medical Officer of Oncology at Lunit, who led the study. “As each immune phenotype leads to different survival rates and response to immune checkpoint inhibitors, it was important to validate that our AI algorithm can accurately predict such classification.”

“Although more and more cancer patients are treated with immune checkpoint inhibitors, prediction of response to them is limited. Novel biomarkers are needed to identify more patients that respond well to this powerful treatment,” said Brandon Suh, CEO of Lunit. “The value that this study presents is in the accuracy of AI-based pathologic classification of immune phenotype, demonstrating the possibility of Lunit SCOPE IO to be utilized as a novel biomarker of immunotherapy in the digital pathology era.”

Based on the company’s commitment in clinical research and studies, Lunit’s findings have also been presented in ASCO and AACR since 2019, acknowledged for its potential value. “We will continue our development as we plan to expand the scope of our analysis and further train the algorithm into a reliable product that can help improve patient outcome,” added Suh.

Abstract Information

Abstract #: 780

Abstract Title: Development and Validation of Deep Learning-Based Pathologic Classification of Immune Phenotype in Non-Small Cell Lung Cancer

Presentation Details: Poster Session on Monday, March 15, 2021 from 2:15 PM - 3:15 PM Pacific Time

USCAP website: https://www.uscap.org/uscap-annual-meeting/