Glioma biopsies Classification Using Raman Spectroscopy and Machine Learning Models on Fresh Tissue Samples.

Area of Science:

• Neuro-oncology
• Optical Spectroscopy
• Biomedical Engineering

Background:

• Accurate identification of glioma cells within normal brain tissue is crucial for complete tumor resection.
• Raman spectroscopy (RS) offers potential for real-time glioma detection, but studies on fresh, untreated tissue are limited.
• Exploring novel Raman bands can enhance the distinction between cancerous and healthy brain tissue.

Purpose of the Study:

• To investigate the efficacy of Raman spectroscopy (RS) on fresh, untreated brain biopsies.
• To identify new Raman bands for differentiating glioma from normal brain tissue.
• To assess the potential of RS as an intraoperative tool for in-vivo glioma detection.

Main Methods:

• Analysis of 63 fresh brain tissue biopsies using Raman spectroscopy shortly after resection.
• Collection and classification of 3450 spectra (1377 Healthy, 2073 Tumor) using machine learning algorithms.
• Screening 135 Raman peaks and utilizing 60 representative peaks for spectral analysis.

Main Results:

• Distinguished between tumor and healthy brain tissue with 83% accuracy and 82% precision.
• Identified 19 new Raman shifts with significant biological relevance.
• Demonstrated the effectiveness of RS in discriminating glioma from healthy brain tissue ex-vivo in fresh samples.

Conclusions:

• Raman spectroscopy is effective and accurate for differentiating glioma from healthy brain tissue in fresh ex-vivo samples.
• The study provides novel spectroscopic data for developing RS into an intraoperative tool for in-vivo glioma detection.
• Further research can leverage these findings to improve real-time surgical guidance for glioma resection.