Advanced Brain Tumor Diagnosis using Haralick Texture Features and Multiclass SVM in MRI Imaging

This study employs contemporary computer techniques and magnetic resonance imaging (MRI) to tackle the critical issue of brain tumor classification. The proposed technique involves obtaining databases, preprocessing the data, extracting features, applying machine learning for classification, and conducting a thorough performance evaluation. The method enhances the visibility of tumors and enhances the characteristics of images by acquiring a diverse collection of MRI images and applying a series of preprocessing operations, including Otsu Binarization, Gray Level Thresholding, Morphological Operations, and Principal Component Analysis (PCA). Subsequently, the Haralick texture characteristics are extracted to depict the textural patterns observed in the pictures. The concept is based on utilizing a multiclass Support Vector Machine (SVM) to achieve accurate tumor classification. Our proposed technique demonstrated efficacy in accurately classifying brain tumors, as evidenced by an impressive accuracy rate of 99.764% and near-perfect ROC curve performance, with area under the curve (AUC) values close to 1. The findings emphasize the potential practicality of the technique and contribute to the advancement of computational methods and medical imaging in the field of brain tumor detection.