Utilizing Transfer Learning for Brain Tumor Detection and Grad-CAM Visual Explanation

Human life is susceptible to a wide range of diseases, from mild conditions to severe threats, with brain tumors posing a particularly significant health challenge. These tumors affect individuals of all ages and contribute substantially to the global disease burden, as reported by the Global Cancer Observatory (GLOBOCAN). Recent advancements in medical image processing, particularly through deep learning technologies, have facilitated faster and more accurate detection of brain cancer. This study explores the application of Convolutional Neural Networks (CNNs) with various transfer learning models, including VGG16, VGG19, EfficientNetB1, EfficientNetB3, EfficientNetB5, and EfficientNetB7. Additionally, the study utilizes Grad-CAM (Gradient-weighted Class Activation Mapping) to provide visual interpretations of the model's predictions, enhancing understanding of the decision-making process. The results indicate that EfficientNetB3 outperforms the other models, achieving an impressive accuracy of 98%. This model has been integrated into a web-based system to assist in the identification of brain tumor images. The combination of EfficientNetB3 and Grad-CAM offers a robust approach for accurate tumor detection, delivering not only high diagnostic performance but also valuable visual insights that can support clinicians in their evaluations.