CNSC-15. DECIPHERING RAMAN-SPECTRAL TISSUE HETEROGENEITY IN GLIOBLASTOMA

Abstract Raman Spectroscopy is able to provide a fast identification method for healthy and pathological tissues without the need sample preparation. This makes it highly interesting intraoperative use in identifying tumor types borders. However, capability of machine-learning classifier recognize known based on spectra relies solid ground truth that usually provided by measuring small tissue samples first analyzing later histopathology. approach limited very closely interspersed cannot be identified macroscopically, such as glioblastoma where vital tissue, necrosis peritumoral with characteristic changes brain are tightly arranged. These areas have been shown distinct their spectral properties, which border aided challenging. As infiltrative tumors leaving no tumor-cell-free borders defined clinical scale, extensive resection must weighed against retaining functional tissue. For this, analysis beyond necessary. In computational we methods unsupervised learning K-means Clustering investigate heterogeneity within glioblastomas native state order reduce complexity this type. Thereby, was possible extract necrotic calculate prediction probability glioblastomas. Several clusters similarity each other could identified, might represent different glioblastoma. Two these were probable grey white matter, respectively, comparison autopsy addition, assignment independently confirmed non-necrotic Deciphering way may great advantage surgeon identify assist control.