Narrow band imaging-based radiogenomics for predicting radiosensitivity in nasopharyngeal carcinoma

Area of Science:

• Oncology
• Radiology
• Genomics

Background:

• Nasopharyngeal carcinoma (NPC) treatment response varies significantly.
• Predicting radiosensitivity is crucial for personalized treatment strategies.
• Current methods for assessing radiosensitivity in NPC require improvement.

Purpose of the Study:

• To evaluate the efficacy of narrow band imaging (NBI) endoscopy-based radiomics for predicting radiosensitivity in NPC.
• To explore the molecular mechanisms underlying NBI radiomics' predictive capability.
• To identify potential biomarkers for treatment response in NPC.

Main Methods:

• Analysis of 267 NBI images from 57 NPC patients using ResNet50 for radiomic feature extraction.
• Application of deep learning and least absolute shrinkage and selection operator (LASSO) for feature selection and gene identification.
• Enrichment analysis and single-cell RNA sequencing to investigate molecular mechanisms and tumor immune microenvironment.

Main Results:

• A radiomics model using 54 selected features achieved high accuracy (0.909) and AUC (0.961) with a random forest algorithm.
• Thirty differentially expressed genes correlated with radiomic features were identified.
• Tumor-associated macrophages were implicated in treatment response, with three key NBI differentially expressed immune genes (NBI-DEIGs)—CCL8, SLC11A1, and PTGS2—identified as regulators.

Conclusions:

• NBI-based radiomics offers a novel and effective method for predicting NPC radiosensitivity.
• The underlying molecular mechanisms likely involve macrophage function, regulated by specific immune genes.
• This approach holds promise for guiding personalized treatment decisions in NPC.