Differentiation of glioblastoma G4 and two types of meningiomas using FTIR spectra and machine learning

Brain tumors are among the most dangerous, due to their location in the organ that governs all life processes. Moreover, the high differentiation of these poses a challenge in diagnostics. Therefore, this study focused on the chemical differentiation of glioblastoma G4 (GBM) and two types of meningiomas (atypical - MAtyp and angiomatous - MAng) were done using Fourier Transform InfraRed (FTIR) spectroscopy, combined with statistical, multivariate, machine learning and rate of spectrum changes methods. The positions of all analyzed peaks differed between GBM and meningiomas. However, for two types of meningiomas, only shift of peaks corresponding to CH₂ bending vibrations, symmetric stretching vibrations of CH_2, amide A, amide I, C=O lipids vibrations, asymmetric stretching vibrations of CH₃ were observed. Principal Component Analysis showed clear differentiation between GBM and the meningiomas. Decision tree clearly showed that wavenumbers corresponding to C=O lipids vibrations provided the highest differentiation between GBM and meningiomas tissues, while amide I for two types of meningiomas. The accuracy and specificity of the results for GBM and meningiomas were more than 90%, while for MAtyp and MAng, these parameters were around 80%.