AI-powered digital scoring of chromogenic in situ hybridization for detection of MET gene amplification

Overview

In this poster, PathAI and AbbVie presented work from a collaborative project to predict MET amplification (MET­amp) status in lung cancer slides where chromogenic in situ hybridization (CISH) was used for analysis of MET gene copy number.

Key Takeaways

• Manual scoring of MET CISH involves pathologists counting MET and control signals in ~50 nuclei, which is time-consuming, subjective, and covers only a small minority of tumor cells on a whole slide image (WSI). In contrast, the AI model exhaustively scored cancer cell nuclei within a CISH WSI for MET amplification.
• The model utilized a hotspot detection algorithm to identify amplification heterogeneity. Use of the hotspot detection model revealed minor foci of MET­amp positivity that would potentially be missed by manual review or by next-generation sequencing analysis, potentially enabling more accurate and sensitive determination of MET­amp status.

USCAP 2026

Partner:  AbbVie