Radiomics- based biomarkers for immunotherapy in non-small cell lung cancer: advances, challenges, and future perspectives in biological interpretation and clinical translation
Abstract
Immune checkpoint inhibitors (ICIs) have improved outcomes in non-small cell lung cancer (NSCLC), yet treatment benefit remains heterogeneous and resistance is common. Genomic alterations, tumor microenvironment (TME) remodeling, and metabolic reprogramming across spatial and temporal scales that are incompletely captured by single-site tissue sampling. Radiomics extracts quantitative features from routine images and offers a framework for investigating macroscopic correlates of these processes. This narrative review critically examines computed tomography (CT) and fluorine-18 fluorodeoxyglucose positron emission tomography/CT (¹8F-FDG PET/CT) radiomics in relation to oncogenic alterations, immune and stromal phenotypes, programmed death-ligand 1 (PD-L1) expression, tumor mutational burden, and metabolic–immune interactions in NSCLC. It should be highlighted that such observed statistical links only generate plausible biological hypotheses and do not constitute experimental proof of true mechanistic relationships. Rather than merely summarizing predictive model performance, we emphasize evidence boundaries and interpretive caveats for radiomic-TME associations. We further elaborate on key translational bottlenecks including limited reproducibility, technical standardization deficits, and retrospective-model overfitting. We discuss ongoing efforts toward cross-scale multi-omics integration, alongside the promises and inherent limitations of explainable artificial intelligence and federated learning for clinical translation. To move beyond purely predictive tools toward mechanism-aware decision-support systems, future radiomic research requires prospective multicenter validation, rigorous data harmonization, and combined analyses integrating spatial transcriptomics, digital pathology, and circulating biomarkers for precision immuno-oncology.
Related
- SPD Matrix Learning for Neuroimaging Analysis: Perspectives, Methods, and Challenges
- Therapeutic cancer vaccines: development, challenges, and future perspectives
- Vesicular Monoamine Transporters at the Crossroads of Structure and Function: Advances, Challenges, and Future Perspectives
- Spatiotemporal remodeling of the tumor–immune interface during neoadjuvant immunotherapy for oral squamous cell carcinoma: pathological response, resistance, and dynamic biomarkers
- Targeting inhibitory phagocytosis checkpoints in cancer immunotherapy: mechanisms, therapeutic strategies, and translational challenges
BibTeX
@article{zhang2026radiomics,
title = {{Radiomics- based biomarkers for immunotherapy in non-small cell lung cancer: advances, challenges, and future perspectives in biological interpretation and clinical translation}},
author = {Ran Zhang and Xiaoqing Ji and Shuran Yang and Zhenlei Zhang and Yian Wang and Jiameng Lu and Degan Lu},
journal = {Frontiers in Immunology},
year = {2026},
month = nov,
doi = {10.3389/fimmu.2026.1871666},
url = {https://www.frontiersin.org/articles/10.3389/fimmu.2026.1871666},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Radiomics- based biomarkers for immunotherapy in non-small cell lung cancer: advances, challenges, and future perspectives in biological interpretation and clinical translation
AU - Ran Zhang
AU - Xiaoqing Ji
AU - Shuran Yang
AU - Zhenlei Zhang
AU - Yian Wang
AU - Jiameng Lu
AU - Degan Lu
PY - 2026
DA - 2026/11/02
JO - Frontiers in Immunology
DO - 10.3389/fimmu.2026.1871666
UR - https://www.frontiersin.org/articles/10.3389/fimmu.2026.1871666
AB - Immune checkpoint inhibitors (ICIs) have improved outcomes in non-small cell lung cancer (NSCLC), yet treatment benefit remains heterogeneous and resistance is common. Genomic alterations, tumor microenvironment (TME) remodeling, and metabolic reprogramming across spatial and temporal scales that are incompletely captured by single-site tissue sampling. Radiomics extracts quantitative features from routine images and offers a framework for investigating macroscopic correlates of these processes. This narrative review critically examines computed tomography (CT) and fluorine-18 fluorodeoxyglucose positron emission tomography/CT (¹8F-FDG PET/CT) radiomics in relation to oncogenic alterations, immune and stromal phenotypes, programmed death-ligand 1 (PD-L1) expression, tumor mutational burden, and metabolic–immune interactions in NSCLC. It should be highlighted that such observed statistical links only generate plausible biological hypotheses and do not constitute experimental proof of true mechanistic relationships. Rather than merely summarizing predictive model performance, we emphasize evidence boundaries and interpretive caveats for radiomic-TME associations. We further elaborate on key translational bottlenecks including limited reproducibility, technical standardization deficits, and retrospective-model overfitting. We discuss ongoing efforts toward cross-scale multi-omics integration, alongside the promises and inherent limitations of explainable artificial intelligence and federated learning for clinical translation. To move beyond purely predictive tools toward mechanism-aware decision-support systems, future radiomic research requires prospective multicenter validation, rigorous data harmonization, and combined analyses integrating spatial transcriptomics, digital pathology, and circulating biomarkers for precision immuno-oncology.
SN - 1664-3224
ER -