ELBP: A Novel Approach for Feature Extraction and Stacking with GLCM in Advanced MRI Brain Tumor Classification

Accurately detecting and classifying brain tumors from MRI imaging is vital in medical image processing, providing valuable insights for disease diagnosis. However, this task’s complexity, significance, and duration remain unchanged. By incorporating a novel Extended Local Binary Patterns (ELBP), Grey Level Co-Occurrence Matrix (GLCM), and data augmentation techniques, a method for image feature extraction technique is proposed which outperforms conventional approaches. By incorporating multiple radii, the ELBP technique provides a more accurate representation of textures than Local Binary Patterns (LBP). GLCM features simultaneously capture higher- order statistical properties of the brain MRI, thereby enhancing the extracted data. The combination of these two feature sets yields an exhaustive image-wide descriptor. Data augmentation such as rotation and rescaling, are utilized to improve the dataset. Chi-squared testing is used to optimize feature selection. In addition to achieving a remarkable 97% accuracy rate, the ensemble classifier demonstrates extraordinary performance across additional evaluation metrics. Findings indicate that combining ELBP and GLCM significantly improves image classification over existing state-of-the-art techniques.