Integrating Haralick Texture Features and Multiclass SVM for Accurate Brain Tumor Diagnosis in MRI Images

Abstract

This research work uses modern computational algorithms and magnetic resonance imaging (MRI) to address the crucial problem of brain tumor categorization. The methodology that has been suggested includes the acquisition of databases, preprocessing, feature extraction, classification based on machine learning, and a comprehensive assessment of performance. The technology improves tumor visibility and refines image features by obtaining a diversified MRI image collection and implementing a sequence of preprocessing operations, such as Otsu Binarization, Gray Level Thresholding, Morphological Operations, and Independent Component Analysis (ICA). Then, in order to describe the textural patterns in the photos, Haralick texture characteristics are retrieved. Using a multiclass Support Vector Machine (SVM) to accurately classify tumors forms the basis of the methodology. Performance evaluation shows the resilience and efficacy of the created classification system through data splitting, metric computation, ROC curve development, cross-validation, and comparisons with current approaches. The findings highlight the methodology's possible practical applicability and advance the field of computational methods and medical imaging intersections for brain tumor identification.