Combination of DWT Variants and GLCM as a Feature for Brain Tumor Classification

Abstract

Brain tumors are a growth of abnormal cells in the intracranial tissue that can disrupt proper brain function. In general, brain tumors are classified into two main categories, benign and malignant. This research aims to classify three types of benign tumors, that are Meningioma (Mg), Glioma (Gl), and Pituitary (Pt) from MRI images. The benign tumors types are classified into four data categories, that are Mg-Gl, Mg-Pt, Gl-Pt, Mg-GI-Pt. The Feature extraction uses Discrete Wavelet Transform (DWT) and Gray Level Co-Occurrence Matrix (GLCM) variant combination as a hybrid feature for recognize and classifying benign tumors types. The classification uses Convolutional Neural Network (CNN) method with ten layers structure. From our experiments, the average accuracy value of DWT combined with four GLCM features, that are Contrast, Homogeneity, Correlation, and Energy is 78.03% in all data categories.