<![CDATA[Diagnosis of brain tumors such as Glioma, Meningioma, and Pituitary using MRI still faces challenges, including reliance on manual interpretation, long evaluation times, and the potential for human error. To address these issues, deep learning-based approaches offer efficient and accurate solutions. This study aims to develop a brain tumor classification model based on deep learning using the InceptionResNetV2 architecture with MixUp augmentation to improve model accuracy and generalization. The model was trained on 7,023 MRI images (Glioma: 1,621; Meningioma: 1,645; Pituitary: 1,757; No-tumor: 2,000), with MixUp proven effective in reducing overfitting and handling data imbalance. The proposed model achieved a highest accuracy of 99.70%, surpassing other models such as CNN with Image Enhancement (97.84%) \[1], Xception (98.00%) \[2], EfficientNet (98.00%) \[3], and ResNet50 (98.47%) \[4]. Evaluation was conducted using metrics including precision, recall, F1-score, as well as MSE, RMSE, and MAE, showing strong performance. These results support the use of transfer learning for medical image classification with limited datasets. This research demonstrates clinical application potential, particularly in improving diagnostic accuracy, speeding up evaluation processes, and reducing human error. Future recommendations include using more diverse datasets, real-world evaluation, and integration into Clinical Decision Support Systems (CDSS).]]>
Copyrights © 2025
