P03.05.b Integrative Radiomic and Immunologic Profiling in Glioblastoma: A Step Toward Imaging-Based Immunophenotyping

Abstract BACKGROUND Glioblastoma is the most common and aggressive form of brain cancer. It is a disease heavily influenced by immunological factors, with immune cells comprising up to one-third of the tumor mass. Apart from immunological studies, non-invasive MRI-based tumor profiling has emerged as a promising field of research. The identification of quantitative imaging biomarkers, also known as radiomics holds promise for enhancing the understanding and characterization of glioblastoma. However, despite the individual importance of both immunology and radiomics imaging in glioblastoma profiling, their combined analysis remains underexplored. MATERIAL AND METHODS A retrospective analysis was conducted on 76 glioblastoma patients enrolled in a prior immunotherapy clinical trial in order to identify immunology and radiomics data at baseline, i.e. before therapy. 34 patients met these criteria. Pre-treatment contrast T1-weighted cerebral magnetic resonance images were analyzed via radiomics and MRI features (including shape, first-order and higher-order features) were extracted. In total, 321 radiomic features were integrated with 67 peripheral blood immunological markers derived from flow cytometry and PCR, enabling comprehensive correlation mapping. RESULTS After excluding redundant radiomic features, in total 324 significant correlations between radiomic imaging and immunological features were discovered. Those imaging variables were further used to build forward- and backward- stepwise regression models, where each immunological parameter was an individual outcome value. We were able to predict 12 immunological features based solely on imaging features. Most interestingly, these were naïve CD8 cells: R2 adjusted=0.540, early differentiated CD8 cells R2 adjusted=0.514 and activated CD 8 cells R2 adjusted=0.532, TH 17 cells R2 adjusted=0.525, TH 2 and TH 17 transcription factor: R2 adjusted=0.708 and 0.688, CD3 from elutra: R2adjusted=0.878. CONCLUSION This study demonstrates that certain immunological parameters, particularly those related to CD8+ and TH cell populations, can be predicted based solely on radiomics imaging data. These findings highlight a potential interplay between complex radiological features and immune profiles in glioblastoma, suggesting that MRI-based radiomics could serve as a non-invasive tool for immunophenotyping and patient stratification in future clinical settings.