Real-time in vivo optical imaging of glioma with confocal laser endomicroscopy in a murine model: comparative analysis of performance of indocyanine green and fluorescein sodium.

OBJECTIVE: Confocal laser endomicroscopy (CLE) enables real-time, cellular-resolution optical biopsies during surgery. Although fluorescein sodium (FNa)-based CLE has been studied extensively, few studies have examined near-infrared (NIR) CLE imaging using indocyanine green (ICG) in brain tumor models. This study compared NIR CLE imaging using intravenous or topical ICG with blue laser CLE using FNa in a mouse glioma model. METHODS: Mice with orthotopically implanted GL261 glioma underwent bilateral craniectomy and CLE imaging. Four mice received intravenous ICG, and 4 received topical ICG. Two mice each received intraperitoneal or topical FNa. CLE imaging was performed using the NIR CLE system and the FNa CLE system on tumor, contralateral normal cortex, and postresection margins and correlated with hematoxylin and eosin (H&E)-stained sections. Signal-to-background ratios (SBRs) were analyzed using an automated 3-step segmentation pipeline and compared between intravenous and topical ICG groups. RESULTS: NIR CLE with intravenous or topical ICG produced a characteristic intracellular fluorescence pattern in tumor tissue, with strong cytoplasmic fluorescence surrounding dark, round nuclei, enabling visualization of cell size, morphology, and pleomorphism consistent with H&E-stained sections. Tumor cells at resection margins were reliably identified by either method. Topical ICG staining of contralateral normal cortex revealed distinct features consistent with normal brain tissue and clearly distinguishable from tumor. Systemic FNa produced a negative-contrast extracellular pattern with tumor cells that appeared as dark silhouettes on a bright background, whereas topical FNa stained tumor and normal tissue with limited morphological differentiation. SBR analysis of intravenous and topical ICG images showed topical ICG produced significantly higher SBRs than intravenous ICG at image level (median 5.7 vs. 4.9, p<0.001) but not animal level. CONCLUSIONS: NIR CLE using ICG provided high-quality cellular-resolution imaging of glioma and normal brain tissue. ICG enabled visualization of cell size and morphology, facilitating clear differentiation between tumor and normal brain. Topical ICG application produced an imaging pattern equivalent to intravenous administration with a similar SBR and allowed visualization of normal brain cytoarchitecture. These findings support ICG for intraoperative NIR CLE imaging of brain tumors and provide a foundation for investigating topical ICG use with CLE imaging.