Imaging Immune Heterogeneity in Nasopharyngeal Carcinoma: A Multidimensional Radiomics Perspective

Nasopharyngeal carcinoma (NPC) is characterized by a distinctive virologic and immunologic profile, in which Epstein-Barr virus-driven immune infiltration coexists with immune escape. This complex interplay gives rise to pronounced immune heterogeneity across spatial, molecular, and temporal dimensions, which critically influences response to immunotherapy and clinical outcomes. However, conventional biopsy-based immune evaluation is limited by sampling bias, invasiveness, and restricted spatial representativeness, precluding comprehensive characterization of the global tumor immune landscape. Radiomics enables high-throughput quantitative extraction of structural, textural, and functional features from routine medical imaging, offering a noninvasive and whole-tumor approach to characterize imaging-derived surrogate phenotypes associated with the tumor immune microenvironment. This review aims to summarize the biologic basis of immune heterogeneity in NPC and to evaluate recent advances in radiomics for noninvasive immune characterization.

We conducted a comprehensive review of recent literature focusing on radiomics-based approaches for assessing the tumor immune microenvironment in NPC. Studies were analyzed from a multidimensional perspective, encompassing spatial characterization of immune-associated imaging patterns, molecular prediction of immune phenotypes such as programmed death-ligand 1 (PD-L1) expression, and predictive modeling of treatment response, together with emerging longitudinal approaches aimed at capturing immune evolution during therapy.

Radiomics enables high-throughput quantitative extraction of structural, textural, and functional features from routine medical imaging, providing a noninvasive and whole-tumor approach to characterize imaging-derived surrogate phenotypes of the tumor immune microenvironment. Accumulating evidence demonstrates its value in capturing spatial immune heterogeneity, predicting molecular immune markers such as PD-L1 expression, and stratifying patients according to treatment response. Additionally, longitudinal radiomics analyses show promise in reflecting dynamic immune evolution during therapy.

Radiomics represents a promising noninvasive tool for comprehensive characterization of immune heterogeneity in NPC. Emerging directions include multimodal imaging integration and deep learning-based virtual immune phenotyping. Addressing key challenges such as standardization, reproducibility, and clinical validation will be essential to enable robust clinical translation.