Enhancing Brain Tumor Detection in MRI with Feature Fusion and SMOTE-Optimization Techniques

Precise detection of brain malignancies using magnetic resonance imaging (MRI) analysis is a significant area in Biomedical image processing. Nevertheless, current methods typically ignore the problem of data imbalance which influences the bias towards the majority class, and employ a single feature extraction strategy, which may have an impact on the classifier’s effectiveness. This study proposes a novel feature fusion method based on intensity based features i.e. First Order Statistics (FOS) and textural based features i.e. Gray Level Co-occurance Matrix (GLCM). The suggested approach employs the Synthetic Minority Over-sampling Technique (SMOTE) to equilibrate the dataset. The merging of features with SMOTE has improved the accuracy of the classifiers. The Grid Search Optimization (GSO) and Grey Wolf Optimization (GWO) are used to optimize the hyperparameters of Classifiers i.e. Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), and Gradient Boosting (GB). Additionally, the GB classifier with GWO technique produces superior results than other optimization technique, with an improved accuracy of 89% and an area under the curve of 0.89. Moreover, the accuracy of the suggested model is 20.38% higher than that of the single feature extraction method. By leveraging the feature fusion and SMOTE-optimization technique, this article aims to achieve accurate classification of brain tumor.