Classification of MRI Brain Tumor with Hybrid VGG19 and Ensemble Classifier Approach

The complex characteristics of brain tumors have prompted the advancement of diverse computer-based categorization methods to enhance the efficiency and precision of MRI diagnoses. However, these processes still needed significant improvement to meet the diagnostic criteria of the actual world. This work aims to improve classification accuracy by combining the powerful feature extraction capabilities of VGG19 with the robustness of ensemble classifiers such as Logistic Regression, Random Forest, and K-nearest neighbor. The suggested technique utilized the Br35H dataset, which comprises 3000 images categorized into tumor-present and tumor-absent cases. The proposed framework underwent evaluation utilizing a 5-fold stratified cross-validation technique, resulting in a mean accuracy of 98.27%. This accuracy outperforms both earlier techniques and traditional deep learning methods. The simplicity and computational cost of the suggested technique indicate its suitability for use in real-world settings with limited resources.