Diagnosis of Glioma Using Raman Spectroscopy and the Entropy Weight Fuzzy-Rough Nearest Neighbor (EFRNN) Algorithm on Fresh Tissue

Abstract

Glioma is the most common tumor of the central nervous system, with high lethality and low cure rates. Glioma is highly infiltrative, making the boundary between glioma and normal brain tissue difficult to determine. Therefore, the accurate tracing of the edge of gliomas is of great importance. Using Raman spectroscopy to distinguish glioma from normal brain tissue has the advantages of speed, noninvasiveness, and non-toxicity. The peak detection method is employed to automatically identify 9 Raman peaks’ characteristic variables including peak position, intensity, and half-wave width, as the variables of the pattern recognition method, which better reflects tissue changes at the molecular level than using the spectrum of one band. 311 Raman spectra of 228 fresh tissues from 196 patients were collected using a portable miniature Raman spectrometer, which is suitable for the rapid intraoperative detection of gliomas. The entropy weight fuzzy-rough nearest neighbor (EFRNN) algorithm was applied for the first time for the discrimination of glioma. The sensitivity of glioma and normal brain tissue classification was 87.21%, the specificity was 86.49%, the positive predictive value was 93.75%, the negative predictive value was 74.42%, and the accuracy was 86.99%. The results demonstrate the potential of applying Raman spectroscopy to the clinical detection of gliomas in vivo and in situ.