2023 World Conference on Lung Cancer (Posters)-EP06.05. Computational Pathology-Based Assessment of cMET IHC Expression for Patient Selection in the Treatment of MET Overexpressing NSCLC

EP06.05. Computational Pathology-Based Assessment of cMET IHC Expression for Patient Selection in the Treatment of MET Overexpressing NSCLC - PDF(Slides)

Back to course

Pdf Summary

The objective of this study was to develop a Quantitative Continuous Scoring (QCS) algorithm for the automated scoring of MET immunohistochemistry (IHC) and identify the optimal feature and threshold associated with improved outcome to treatment of EGFRm NSCLC with savolitinib in combination with osimertinib in patients progressing on prior osimertinib. The study demonstrated good performance of the algorithm and strong correlation with conventional pathologist scoring. High MET expression was observed in patients progressing on osimertinib, confirming one of the most common resistance mechanisms. The study also showed a significant association of MET expression quantified by QCS and patient outcome. A QCS signature was optimized in the SAVANNAH clinical trial, which can identify patients with improved response to treatment with savolitinib plus osimertinib. The combination therapy will be further investigated in the SAFFRON clinical trial. The QCS-based quantification of MET expression showed correlation with conventional scoring, association with efficacy, and identified a representative feature and threshold to identify patients with an increased likelihood to benefit from treatment. The study utilized deep learning algorithms for image analysis and utilized a Hue-Saturation-Density model for stain recognition in digital images. The computational pathology-based assessment of cMET IHC expression showed promise for patient selection in the treatment of MET over-expressing NSCLC. The QCS system generated multiple readouts based on stain intensity, and the quantile 90 of membrane OD emerged as one of the best performing features to stratify patients by efficacy. The study compared visual scoring with the QCS score and showed similar results for response, survival, and prevalence. Further optimization can lead to the identification of different QCS readouts for better patient stratification.

Asset Subtitle

Simon Christ

Meta Tag

Speaker Simon Christ

Topic Pathology & Biomarkers

Keywords

Quantitative Continuous Scoring

MET immunohistochemistry

EGFRm NSCLC

savolitinib

osimertinib

resistance mechanisms

SAVANNAH clinical trial

SAFFRON clinical trial

deep learning algorithms

computational pathology