Descriptive Statistical Analysis on Intensity and Texture Features to Compare Hyperspectral Images of Unstained Tissues Samples

This study explored the use of hyperspectral microscopy to differentiate between two grades of glioma tumors. Both spatial and spectral information from hyperspectral data-cube were analyzed to extract distinguishing features. The ability to separate the tumor classes was evaluated with descriptive statistical tools across selected wavelength intervals. We found that several spectral (intensity, slope) and texture (Contrast, Correlation, Energy, Homogeneity) features show distinct, wavelength-dependent variations between the two tissue classes, enabling their effective separation. It is noteworthy that, this differentiation between low- and high-grade tumors was accomplished on unstained samples. This study demonstrates that, by analyzing a sample's intrinsic spectral response, which correlates with its biochemical structure and metabolism-related processes, hyperspectral microscopy could significantly increase the diagnostic accuracy in Pathology, holding a real potential for rapid diagnosis with reduced chemical use.