2023 World Conference on Lung Cancer (Posters)-P1.22. Immediate Biomarkers Profiling of Actionable NSCLC Panel from a Digitized H&E Biopsy Image

P1.22. Immediate Biomarkers Profiling of Actionable NSCLC Panel from a Digitized H&E Biopsy Image - PDF(Slides)

Pdf Summary

The document discusses the performance of an AI-based biomarker panel called LungOI for non-small cell lung cancer (NSCLC). The results show that in high-confidence cases, the biomarker panel had an average area under the curve (AUC) of 0.956, sensitivity of 93.9%, specificity of 94.7%, and accuracy of 94.7%.<br /><br />The performance of the biomarker panel for different biomarkers was also evaluated. ALK fusion had an AUC of 0.979, sensitivity of 0.974, specificity of 0.962, and accuracy of 0.962. BRAF V600E had an AUC of 0.983, sensitivity of 0.941, specificity of 0.973, and accuracy of 0.973. EGFR mut had an AUC of 0.973, sensitivity of 0.901, specificity of 0.917, and accuracy of 0.914. ERBB2 mut had an AUC of 0.990, sensitivity of 1.000, specificity of 0.973, and accuracy of 0.973. MET ex14 skipping had an AUC of 0.946, sensitivity of 0.952, specificity of 0.921, and accuracy of 0.923.<br /><br />The training and testing of the biomarker panel were performed on whole slide images (WSIs) of lung cancer samples using self-supervised learning algorithms. The model's performance was evaluated on an independent lung cancer dataset and compared to the results of reference methods such as next-generation sequencing (NGS) and fluorescence in situ hybridization (FISH).<br /><br />For each sample, the biomarker panel assigned a positive or negative status to each gene based on high-confidence biomarker scores, or an inconclusive status for low-confidence scores.<br /><br />The document highlights the use of the AI-based solution, which analyzes H&E-stained slides, in detecting actionable genomic alterations in NSCLC compared to clinically diagnostic tests.

Asset Subtitle

Nurit Paz-Yaacov

Meta Tag

Speaker Nurit Paz-Yaacov

Topic Pathology & Biomarkers: Genetic Biomarkers

Keywords

AI-based biomarker panel

LungOI

non-small cell lung cancer

performance

sensitivity

specificity

accuracy

biomarkers

whole slide images

actionable genomic alterations