Abstract
Purpose
Raman spectroscopy (RS) is a promising method for brain tumor detection. Near-infrared autofluorescence (AF) acquired during RS provides additional information useful for tumor identification and was investigated in comparison with RS for delineating brain tumors in situ.
Methods
Raman spectra together with AF were acquired in situ within the solid tumor and at the tumor border during routine brain tumor surgeries (218 spectra; glioma WHO II-III n = 6, GBM n = 10, metastases n = 10, meningioma n = 3). Tissue classification for tumor identification in situ was trained on data acquired ex vivo (375 spectra; glioma/GBM patients n = 20, metastases n = 11, meningioma n = 13, epileptic hippocampi n = 4).
Results
Both in situ and ex vivo data showed that AF intensity in brain tumors is lower compared to border regions and normal brain tissue. Moreover, a positive correlation was observed between the AF intensity and the intensity of the Raman band of lipids at 1437 cm− 1, while a negative correlation was found with the intensity of the protein band at 1260 cm− 1. The classification of in situ AF and RS datasets matched the surgeon’s evaluation of tissue type with a correct rate of 0.83 and 0.84, respectively. Similar correct rates were achieved in comparison to histopathology of tissue biopsies resected in selected measurement positons (AF: 0.80, RS: 0.83).
Conclusions
Spectroscopy was successfully integrated into existing neurosurgical workflows and in situ spectroscopic data could be classified based on ex vivo data. RS confirmed its ability in detecting brain tumors, while AF emerged as a competitive method for intraoperative tumor delineation.