Abstract:
Glioma is the most common and aggressive kind of primary brain tumour, representing 16% of all neoplasms in the brain and central nervous system (CNS). Gliomas are categorized into four grades (I, II, III, and IV) by the World Health Organization (WHO) based on the tumors' aggressiveness. Gliomas in grades I and II are categorized as low-grade, and grades III and IV are high-grade. Glioblastoma (grade IV) is the most aggressive and dangerous of all the others. The method of magnetic resonance imaging (MRI) is commonly utilized for tumour diagnosis and categorization. Classifying MR images manually is time-consuming and difficult. Automatic or semiautomatic classification methods are necessary to differentiate between different types of tumors because human observations may lead to errors in classification. In this study, low-grade gliomas (LGG) and glioblastomas (GBM) were differentiated using MRI, and their textural properties were examined using the Gray Level Co-occurrence Matrix (GLCM). Axial images were selected using the MicroDicom viewer T2- FLAIR (fluid-attenuated inversion recovery). The active, oedema, and full tumour (with oedema) regions of ROI (Region of Interest) were drawn for selected image slices using MATLAB. The texture features that are based on normalized GLCM are contrast, correlation, energy, and homogeneity. Each ROI’s feature values were obtained independently. The t-test was used to compare the LGG feature values with the GBM feature values for the differentiation of tumour types. Feature values for the oedema and the entire tumour region were statistically insignificant due to p-values being higher than 0.05, but for the active region, the p-values corresponding to contrast, correlation, and homogeneity were 0.0092, 0.0217, and 0.0092, respectively. Hence, these feature values were statistically significant. According to this study, it was possible to distinguish GBM (Glioblastoma) from LGG (Low-Grade Gliomas) using contrast, correlation, and homogeneity in the active region.
