Glio-SERS: Artificial intelligence and surface enhanced raman spectroscopy driven liquid biopsy method for brain tumor classification

Glioblastoma (GB), the most aggressive adult brain tumor, requires invasive procedures such as biopsy or surgical intervention, along with sophisticated laboratory settings and prolonged, costly molecular testing for accurate diagnosis. Nearly all patients with GB experience tumor regrowth within two years after primary surgery. The current management of gliomas relies on diagnostic imaging, which lacks the high sensitivity and specificity needed to evaluate recurrence after primary treatment. Implementing liquid biopsy in gliomas is essential for early diagnosis, detecting residual disease after surgery, and assessing disease status post-treatment. With an average survival rate of just 14 months, there is an urgent need for rapid, accurate, cost-effective, and minimally invasive diagnostic strategies. This study introduces a transformative diagnostic paradigm—a liquid biopsy approach that merges Surface-Enhanced Raman Spectroscopy (SERS) analysis of exosomes with artificial intelligence (AI). In a prospective study, we evaluated the efficacy of this SERS and AI-based liquid biopsy analysis for GB detection. We collected 20 glioblastoma, 24 meningioma (MNG) as the most frequent benign tumor control, and 30 healthy control (HC) plasma samples under informed consent and IRB approval. SERS was employed to analyze the isolated plasma exosomes, generating spectral data on molecular signatures indicative of each condition. Deep learning algorithms were integrated into the analysis pipeline to facilitate rapid and accurate differentiation between GB, MNG, and HC plasma exosome SERS signatures. Our approach achieved a remarkable 87\% prediction accuracy in distinguishing GB exosomal signatures from those of MNG and HC individuals. This result signifies a substantial advancement in the precision and speed of GB diagnostics compared to traditional methods. This pioneering liquid biopsy technique emerges as a front-line solution for GB detection, differentiation, and monitoring. The ongoing integration of an expansive library of tumor signatures signifies a significant leap forward in analytical technologies tailored for neuro-oncology. Our study demonstrates not only the technological superiority of our approach but also its potential to revolutionize GB detection and improve patient care once it is clinically validated with a larger cohort.