A Comparative Analysis of EfficientNetB0 and EfficientNetV2 Variants for Brain Tumor Classification Using MRI Images

Accurate and early diagnosis of brain tumors is critical for effective treatment planning, yet traditional methods of analyzing Magnetic Resonance Imaging (MRI) scans are labor-intensive and prone to variability among experts. Deep learning, particularly Convolutional Neural Networks (CNNs), has emerged as a transformative tool in medical imaging by automating feature extraction and enhancing classification accuracy. This study provides a comparative analysis of EfficientNetB0 and three EfficientNetV2 variants (S, M, and L) for brain tumor classification using the Figshare Brain Tumor Dataset, which includes glioma, meningioma, and pituitary tumors. Each model was evaluated using metrics such as accuracy, precision, recall, F1-score, and ROC-AUC. The results reveal that EfficientNetV2-S outperformed other models, achieving the highest accuracy of 98.20% and delivering balanced performance across all classes. EfficientNetV2-M and EfficientNetV2-L also demonstrated strong classification capabilities, with minor trade-offs in computational efficiency. These findings highlight the potential of EfficientNetV2 architectures for automated and reliable brain tumor classification, offering significant advantages for clinical applications. Future work could focus on integrating multi-modal imaging data and optimizing models for deployment in real-time diagnostic settings.