|
Gliomas are extremely infiltrative type of brain cancers, the borders of which are difficult to locate. Gliomas largely
consist of tumors of astrocytic or oligodendroglial lineage. Usually stereotactic surgery is performed to obtain tumor
tissue sample. Complete excision of these tumors with preservation of uninvolved normal areas is important during
brain tumor surgeries. The present study was undertaken to explore feasibility of classifying abnormal and normal
glioma tissues with Raman spectroscopy (RS). RS is a nondestructive vibrational spectroscopic technique, which
provides information about molecular composition, molecular structures and molecular interactions in tissue. Postoperated
33 (20-abnormal and 13-normal) gliomas tissue samples of different grades were collected under clinical
supervision. Five micron section from tissue sample was used for confirmatory histopathological diagnosis while the
remaining tissue was placed on CaF2 window and spectra were acquired using a fiberoptic-probe-coupled HE-785
Raman-spectrometer. Spectral acquisition parameters were laser power-80mW, integration-20s and averaged over 3
accumulations. Spectra were pre-processed and subjected to unsupervised Principal-Component Analysis (PCA) to
identify trends of classification. Supervised PC-LDA (Principal-Component-Linear-Discriminant Analysis) was
used to develop standard-models using spectra of 12 normal and abnormal specimens each. Leave-one-out crossvalidation
yielded classification-efficiency of 90% and 80% for normal and abnormal conditions, respectively.
Evaluation with an independent-test data-set comprising of 135 spectra of 9 samples provided sensitivity of 100%
and specificity of 70%. Findings of this preliminary study may pave way for objective tumor margin assessment
during brain surgery.
|
