Explainable Brain Tumor Classification Using VGG16 and Grad-CAM on MRI Scans

This research introduces an explainable deep learning framework that uses the VGG16 architecture with transfer learning to classify brain tumors. A carefully selected MRI dataset with four categories—glioma, meningioma, pituitary tumor, and no tumor—is used to fine-tune the model. Since deep learning achieves outstanding classification performance and allows for automatic feature extraction,medical image analysis has drastically changed. However, it takes a lot of resources and vast labeled datasets to train sophisticated networks from the start. This is addressed by the suggested method, which uses a pre-trained VGG16 model that was initially learned on ImageNet and fine-tunes it for brain MRI classification. Accuracy, precision, recall, and F1-score are utilized to evaluate the model’s performance; Gradient-weighted Class Activation Mapping (Grad-CAM) contributes to additional interpretability. The results offer visual explanations for the model’s predictions and demonstrate that it executes consistently, particularly in the “No Tumor” and “Pituitary Tumor” classes. The study illustrates how explainable AI methods and transfer learning could potentially be inte- grated to improve diagnostic help and transparency in applications for medical imaging.