CNSC-17. RAPID RAMAN DISCRIMINATION OF GLIOBLASTOMA EDEMA, INFILTRATION, AND ENHANCEMENT VIA SINGLE-CELL AND SPATIAL OMICS

Abstract INTRODUCTION In glioblastoma (GBM), extending resection beyond the MRI-enhancing margin to include infiltrative regions is recommended for improved outcomes; no objective biomarker for infiltration exists. Raman spectroscopy rapidly provides label-free molecular information, offering potential for intraoperative boundary assessment. We investigated whether Raman spectra could distinguish edematous, infiltrative, and contrast-enhancing GBM regions and integrated single-cell Raman with transcriptomics (Xenium®) to elucidate underlying molecular mechanisms. METHODS Seventeen GBM patients undergoing resection from June 2023 to March 2025 were enrolled. Based on MRI findings, navigation-guided stereotactic sampling was performed in edematous, infiltrative, and contrast-enhancing regions. Normal white matter obtained from epilepsy surgery served as a control. Tissue samples were measured using a 532 nm excitation Raman microscope, then classified via principal component analysis and support vector machine. Single-cell suspensions were prepared for Raman measurement and clustering. Spatial omics analysis was performed on normal white matter and contrast-enhancing regions to validate the spectral findings. RESULTS We obtained 17 enhancing, 8 infiltrative, 7 edematous, and 11 control samples. Tissue-level spectra showed a significant increase in carotenoid peaks (1156, 1514 cm^-1) and a decrease in lipid peaks (1440, 1654 cm^-1) along the edematous–infiltrative–enhancing gradient (p<0.05). Five-fold cross-validation yielded classification accuracies of 96.7% (normal), 71.5% (edema), 81.5% (infiltration), and 93.0% (enhancement). Single-cell Raman analyses identified three clusters: normal cells, tumor cells, and tumor-associated macrophages (TAMs). TAMs exhibited unique spectral profiles and contributed to tissue-level spectral differences. Spatial omics confirmed upregulation of lipid transport genes in TAMs, supporting the Raman findings. CONCLUSION Raman spectroscopy enables accurate discrimination of edematous, infiltrative, and contrast-enhancing GBM regions at both tissue and single-cell levels. Integration with spatial transcriptomics reveals the molecular basis of spectral variations, involving TAM-mediated lipid metabolism. This approach may facilitate intraoperative delineation of tumor margins and enhance our understanding of GBM infiltration mechanisms.