Publication

Abstract : Accurate classification and evaluation of brain tumor severity are crucial for prompt diagnosis, treatment strategy, and enhanced patient results. Deep learning techniques, especially Convolutional Neural Networks (CNNs) and transfer learning methods, have recently become prominent in medical imaging to automate the analysis of brain tumors. This research examines how three models-ResNet50, EfficientNetB0, and a custom CNN-perform on a standard brain tumor MRI dataset. The assessment reveals that although ResNet50 attained a modest validation accuracy of 56.88%, EfficientNetB0 and the CNN exceeded it with validation accuracies of 77.56% and 80.02%, respectively. These results illustrate the capability of efficient and robust architectures in managing intricate medical imaging challenges. The proposed work highlights the importance of deep learning in both identifying tumor types and evaluating their severity, confirming the clinical standards for tumor grading. The findings indicate that upcoming studies need to concentrate on integrating hybrid deep learning models, increasing dataset variety, and utilizing explainable AI methods to boost interpretability and clinical confidence