2024 Asia Conference on Lung Cancer (ACLC) - Poster & ePosters-PP01.04

PP01.04 - Sunhee Chang

Pdf Summary

This study investigates the interobserver agreement of MET immunohistochemistry (IHC) interpretation in non-small cell lung cancer (NSCLC) among 33 pathologists and between pathologists and an artificial intelligence (AI) model. MET, or mesenchymal-epithelial transition factor, is often overexpressed in NSCLC. Accurate identification through IHC is crucial for targeted therapies. Recent clinical trials have shown promising results with treatments directed at MET overexpression, but interobserver variability in IHC interpretation remains a challenge.<br /><br />The study involved 30 lung adenocarcinoma cases from Samsung Medical Center, evaluated using the SP44 antibody and scored by the Ventana BenchMark ULTRA IHC/ISH system. Thirty-three pathologists assessed the whole slide images based on a scoring system where scores were given for no (0), weak (1), moderate (2), or strong (3) staining. An AI model, DeepBio, was also utilized for scoring. H-scores were derived from the staining scores, with positive results defined as 50% strong staining cells or H-score greater than or equal to 150.<br /><br />The results showed moderate to substantial agreement among pathologists, with the highest concordance observed among more experienced pathologists and for biopsy specimens. The AI model demonstrated moderate to perfect agreement with pathologists, offering a consistent approach that might augment conventional interpretation methods.<br /><br />In conclusion, while pathologist interpretation of MET IHC showed reasonable agreement, experience and specimen type influenced outcomes. The AI model provided comparable results, suggesting its potential utility in standardizing MET IHC interpretation, which could enhance accuracy and consistency in clinical settings. This study underscores the potential for AI to mitigate variability in MET IHC interpretation, potentially improving the precision of diagnostics and patient outcomes in NSCLC therapy.

Keywords

MET

IHC

NSCLC

interobserver agreement

artificial intelligence

pathologists

lung adenocarcinoma

SP44 antibody

DeepBio model

diagnostics