Evaluation of MRI Radiomic Features as Surrogates of Tumor Immune Status in Breast Cancer: A Pilot Study of MRI-Derived Radiomic Features for Non-Invasive Immune Profiling in Breast Cancer.
- 1Irma Lerma Rangel College of Pharmacy, Texas A&M University Health Science Center, Kingsville, TX 78363, USA.
- 2Veterans Memorial High School, Corpus Christi, TX 78414, USA.
Abstract
Background
Tumor-infiltrating lymphocytes (TILs) and CD8+ T cell density are important predictors of immunotherapy response in breast cancer, but their evaluation requires invasive tissue biopsies. Radiomics, which enables high-throughput extraction of quantitative imaging features, offers a promising non-invasive approach to characterize the tumor immune microenvironment (TIME). This study evaluates whether MRI-derived radiomic features could predict immune infiltration status in breast cancer. Additionally, we have developed an on-line browser-based tool that can predict the immune status of the tumor environment.Methods
Preoperative MRI and matched transcriptomic data from 38 breast cancer patients were obtained from TCIA and TCGA-BRCA datasets. Tumors were manually segmented using 3D Slicer, and radiomic features were extracted using PyRadiomics. Eighteen biologically relevant radiomic features associated with immune status were selected for modeling. Immune markers included CD8B gene expression, CD8+ T cell infiltration estimated by CIBERSORT, and Immune Score estimated by ESTIMATE. Machine learning models, including Random Forest, Support Vector Machine, Deep Learning, and Elastic Net, were trained to classify immune-high versus immune-low tumors, followed by validation in an independent cohort of 13 patients.Results
Random Forest demonstrated strong internal performance (AUCs of 0.945, 0.993, and 0.974 for CD8B expression, CD8+ T cell infiltration, and Immune Score, respectively), but this dropped substantially in the held-out test partition and in an independent external validation cohort (n = 13, AUC 0.44-0.64), indicating that the internal metrics likely reflect overfitting at this sample size and should not be interpreted as evidence of generalizable performance. Additionally, we developed RadioImmune, a browser-based prototype research/demo tool that uses a logistic-regression approximation (not the Random Forest model above) for exploratory, real-time estimation of total immune score and CD8+ T cell infiltration status from radiomic inputs; its performance is not represented by the Random Forest AUCs reported above.Conclusions
These findings provide proof-of-concept that MRI-based radiomic features may serve as potential non-invasive biomarkers of the breast tumor immune microenvironment, pending validation in larger, independent cohorts.
Related
- CT radiomic features detect mineral, collagen, and water contributions to cortical bone texture
- Hybrid deep learning and meta learning for Alzheimer disease classification via multimodal radiomic and visual features
- Effect of Intensity Discretization on Radiomic Features in En Face EDI-OCT of Choroidal Nevi
- Lobar analysis of classical and deep learning radiomic features in chronic airflow limitation
- OCT-Derived Radiomic Features for Differentiating Optic Neuritis Etiologies
BibTeX
@article{shah2026evaluation,
title = {{Evaluation of MRI Radiomic Features as Surrogates of Tumor Immune Status in Breast Cancer: A Pilot Study of MRI-Derived Radiomic Features for Non-Invasive Immune Profiling in Breast Cancer.}},
author = {Mimansha Shah and Soaham Kumar and Narendra Kumar and Jayshree Mishra},
journal = {Cancers},
year = {2026},
month = sep,
doi = {10.3390/cancers18193087},
url = {https://europepmc.org/article/MED/42882782},
note = {CC BY},
}
RIS
TY - JOUR
TI - Evaluation of MRI Radiomic Features as Surrogates of Tumor Immune Status in Breast Cancer: A Pilot Study of MRI-Derived Radiomic Features for Non-Invasive Immune Profiling in Breast Cancer.
AU - Mimansha Shah
AU - Soaham Kumar
AU - Narendra Kumar
AU - Jayshree Mishra
PY - 2026
DA - 2026/09/23
JO - Cancers
DO - 10.3390/cancers18193087
UR - https://europepmc.org/article/MED/42882782
AB - Background Tumor-infiltrating lymphocytes (TILs) and CD8+ T cell density are important predictors of immunotherapy response in breast cancer, but their evaluation requires invasive tissue biopsies. Radiomics, which enables high-throughput extraction of quantitative imaging features, offers a promising non-invasive approach to characterize the tumor immune microenvironment (TIME). This study evaluates whether MRI-derived radiomic features could predict immune infiltration status in breast cancer. Additionally, we have developed an on-line browser-based tool that can predict the immune status of the tumor environment.Methods Preoperative MRI and matched transcriptomic data from 38 breast cancer patients were obtained from TCIA and TCGA-BRCA datasets. Tumors were manually segmented using 3D Slicer, and radiomic features were extracted using PyRadiomics. Eighteen biologically relevant radiomic features associated with immune status were selected for modeling. Immune markers included CD8B gene expression, CD8+ T cell infiltration estimated by CIBERSORT, and Immune Score estimated by ESTIMATE. Machine learning models, including Random Forest, Support Vector Machine, Deep Learning, and Elastic Net, were trained to classify immune-high versus immune-low tumors, followed by validation in an independent cohort of 13 patients.Results Random Forest demonstrated strong internal performance (AUCs of 0.945, 0.993, and 0.974 for CD8B expression, CD8+ T cell infiltration, and Immune Score, respectively), but this dropped substantially in the held-out test partition and in an independent external validation cohort (n = 13, AUC 0.44-0.64), indicating that the internal metrics likely reflect overfitting at this sample size and should not be interpreted as evidence of generalizable performance. Additionally, we developed RadioImmune, a browser-based prototype research/demo tool that uses a logistic-regression approximation (not the Random Forest model above) for exploratory, real-time estimation of total immune score and CD8+ T cell infiltration status from radiomic inputs; its performance is not represented by the Random Forest AUCs reported above.Conclusions These findings provide proof-of-concept that MRI-based radiomic features may serve as potential non-invasive biomarkers of the breast tumor immune microenvironment, pending validation in larger, independent cohorts.
SN - 2072-6694
ER -