HYBRID DEEP LEARNING AND MACHINE LEARNING APPROACH FOR BRAIN TUMOR DETECTION

Abstract

The accurate and early diagnosis of brain tumors is critical due to their high mortality rates. Although Magnetic Resonance Imaging (MRI) is the primary diagnostic modality its manual interpretation is often subjective. This paper introduces an automated hybrid framework for brain tumor classification that merges deep transfer learning with classical machine learning. The approach employs DenseNet121 for initial feature extraction, augmented by a Convolutional Block Attention Module (CBAM) to enhance discriminative features and suppress noise. The resulting feature vectors are classified using multiple algorithms, including Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), K-Nearest Neighbor (KNN) and XGBoost. Empirical results demonstrate the framework's high performance, with Logistic Regression achieving a peak accuracy of 94%. However, XGBoost proved to be the most balanced and effective model, attaining a precision of 95.35% and a recall of 95.03%, establishing it as the optimal classifier for this task while LR presents a robust alternative for efficient deployment.