Abstract 1027: A robust predictive marker of clinical outcomes to immune-checkpoint inhibition in advanced squamous head-and-neck cancer calculated from histopathological slides

Key Points

• The aim is to develop an accurate biomarker for predicting clinical outcomes of immune-checkpoint inhibition in HNSCC using histopathological slides.
• High-resolution scans of pre-treatment tumor H&E slides from 89 advanced HNSCC patients were obtained.
• A model was trained using leave-4-out cross-validation to predict response rates and survival.
• Tile features were extracted from slides using deep learning and aggregated using attention-based learning.
• The model achieved a ROC AUC of 0.67 for predicting overall response rate in the NYCU cohort.
• In the Hadassah cohort, the model showed ROC AUC of 0.76 and significant stratification of progression free survival.
• CPS demonstrated no predictive power in any cohort, while ENLIGHT-DP consistently outperformed CPS.