<![CDATA[Brain tumors are one of the most deadly neurological diseases that require early and accurate diagnosis to determine the right treatment plan. The use of Magnetic Resonance Imaging (MRI) images is the standard in detecting brain tumors, but manual classification by radiologists is time-consuming and has a high risk of subjectivity. This study focuses on the classification of four main categories: glioma, meningioma, pituitary tumor, and healthy brain tumor (not a tumor). This study aims to build an automatic multi-class classification system for brain tumors using a Deep Learning approach with MobileNetV2 and EfficientNetB0 architectures. The training process is carried out using transfer learning techniques and learning rate optimization through system callbacks to ensure the model reaches the best convergence point. The results show that the proposed model is capable of classification with very high performance, achieving an accuracy of 96.88%. The evaluation results using a confusion matrix indicate that the model has a consistent ability to distinguish between tumor classes with an average F1 score of 0.97.]]>
Copyrights © 2026
