Akademik Veri Yönetim Sistemi
23th INTERNATIONAL İSTANBUL SCIENTIFIC RESEARCH CONGRESS ON LIFE, ENGINEERING, ARCHITECTURE, AND MATHEMATICAL SCIENCES, İstanbul, Türkiye, 20 - 22 Kasım 2025, ss.530-538, (Tam Metin Bildiri)
Brain tumors are among the most complex and life-threatening diseases of the central nervous system, where early and accurate diagnosis plays a critical role in improving clinical outcomes and extending patient survival. Magnetic Resonance Imaging (MRI) is one of the most commonly used methods for evaluating the anatomical and pathological characteristics of brain tissue due to its high-resolution and non-invasive nature. However, manual interpretation of MRI images is a time-consuming and subjective process that is prone to human error. Therefore, this study proposes a comprehensive deep learningbased approach for the automatic classification of brain tumors from MRI images. The study utilized the Brain Tumor MRI Dataset available on the Kaggle platform, consisting of 7,023 MRI images categorized into four classes: glioma, meningioma, pituitary, and no tumor. The images were preprocessed through normalization, grayscale conversion, resizing, and data augmentation to enhance the model’s generalization capability. Several modern convolutional neural network (CNN) architectures, including EfficientNetB0, MobileNetV2, and VGG16 were trained and compared. In addition, a lightweight custom CNN model was designed to achieve high accuracy while minimizing computational cost. All models were trained using the Adam optimization algorithm and evaluated through 5-fold cross-validation to ensure reliability. The experimental results demonstrated that the proposed deep learning approach achieved an average accuracy of over 92.72% (±2.91), outperforming most of the pre-trained models. The combination of transfer learning, dropout, and data augmentation techniques enhanced model robustness and significantly reduced overfitting. In conclusion, this study shows that deep learning methods can serve as a reliable decision-support tool for assisting radiologists in the accurate and early classification of brain tumors.