SFTNet: A Hybrid Approach Combining Deep Learning and Machine Learning for Brain Tumor Classification

Brain Tumors (BTs), characterized by the abnormal proliferation of cells, cause significant damage to vital blood vessels and nerves in the body. Misdiagnosis can result in inappropriate medical interventions, reducing patient survival rates. Therefore, accurate and timely diagnosis is crucial for improving outcomes. This study explores an integrated approach combining deep learning (DL) and traditional machine learning (ML) techniques for BT detection and classification. While conventional ML methods require labor-intensive handcrafted features, DL offers robust automatic feature extraction, gaining prominence in classification tasks. In this study, various experiments were conducted, integrating ResNet-50, MobileNetV2, EfficientNetB0, and a Custom CNN (CCNN) with support vector machines (SVM) and K-nearest neighbors (KNN) for tumor classification. The combination of DL and ML involved feature extraction using ResNet-50, MobileNetV2, EfficientNetB0, and CCNN, followed by classification with SoftMax, SVM, and KNN. The Brain Tumor MRI dataset, containing 3,264 images across three tumor types and one normal class, was used for this research. MRI images were enhanced using Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve feature quality. All proposed methods showed strong performance, with the hybrid SFTNet (CCNN+SVM) model achieving an impressive 97% accuracy and a 97% recall across the tumor classes, alongside a near-perfect 99% for the normal class.