Pathomics-based machine learning model predicts interferon-gamma expression in head and neck squamous cell carcinoma...

Area of Science:

• Oncology
• Immunology
• Computational Pathology

Background:

• Interferon-gamma (IFNG) is crucial for immune responses in head and neck squamous cell carcinoma (HNSCC).
• IFNG influences the efficacy of immune checkpoint inhibitors in HNSCC treatment.
• Understanding IFNG expression is vital for HNSCC pathophysiology and therapeutic strategies.

Purpose of the Study:

• To develop a machine learning model predicting IFNG expression using histopathological images in HNSCC.
• To analyze the tumor immune microenvironment in HNSCC through pathomics.
• To establish a link between IFNG expression, pathomics, and patient survival outcomes.

Main Methods:

• Utilized 271 TCGA-HNSCC cases and 71 hospital validation cases.
• Applied OTSU algorithm for tumor segmentation and PyRadiomics for feature extraction (1,488 features).
• Employed mRMR and RFE for feature selection, constructing a Gradient Boosting Machine (GBM) model.

Main Results:

• The GBM model achieved strong survival prediction performance (AUCs: 0.836 training, 0.753 validation, 0.740 hospital).
• Distinct pathway activation patterns were observed, with immune evasion pathways prominent in high PS subgroups.
• Elevated IFNG expression correlated with longer median survival in HNSCC patients.

Conclusions:

• Pathomics analysis of histopathological images can effectively predict IFNG expression in HNSCC.
• The correlation between IFNG and pathomics serves as a biomarker framework for HNSCC.
• This approach may enable non-invasive characterization of the immune microenvironment for personalized HNSCC therapies.