Fresh Brain Tissue Diagnostics Using Raman Spectroscopy In Humans

A key part of treatment for brain tumors is tumor resection, the amount of tumor resection has a direct correlation on the mortality of the patient. However, tools available to surgeons to ensure maximal resection while preserving a patient's functional status often increase surgery time or are lacking in accuracy. Raman spectroscopy is a promising modality that can provide real time intraoperative data to the surgeon about whether a tissue is normal or cancerous, thus aiding the surgeon in obtaining maximal safe resection. In this paper, we pair Raman spectroscopy with a supervised machine learning technique, support vector machine, to classify fresh tissue samples as solid tumor, infiltrating tumor, necrosis, or normal brain tissue. The support vector machine classifier resulted in 89% accuracy of the 117 tissue samples while delivering real time results. Accurate real-time tissue characterization will vastly reduce the amount of time a surgeon waits for the results of biopsies and provides a means for intraoperative margin assessment during tumor resection.