MRI-Based Brain Tumor Classification with Deep Learning Methods
Bildiriler > Tam Metin Bildiri
Yazarlar: Melike Kılıç, Seyit Metin Köroğlu, Duygu Işık, Savaş İZZETOĞLU, Züleyha AKUSTA DAĞDEVİREN
2025 9th International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT), , , Uluslararası, Tam Metin Bildiri, ss 1-6
DOI: 10.1109/ismsit67332.2025.11268236
Özet
In this study, a custom Convolutional Neural Network (CNN) model and EfficientNetB0 (a transfer learning model), were comparatively evaluated with Magnetic Resonance Imaging (MRI) data to classify brain tumors. Brain tumors are of critical importance due to their profound impact on the central nervous system. Accurate and early diagnosis is an important factor that plays a valuable role in creating effective treatment strategies. The study addressed four types of brain tumors: pituitary, meningioma, glioma and no tumor. Initially, a custom CNN model was developed, yielding validation and test accuracies of 65% and 54%, respectively, which indicate limited classification performance. Subsequently, the EfficientNetB0 model was employed through a transfer learning approach, achieving substantially higher validation accuracy (0.96), test accuracy (0.97), and improved generalization capability. The findings clearly demonstrate that EfficientNetB0 surpasses the custom CNN model in brain tumor classification performance. These results indicate that EfficientNetB0 shows strong potential for a reliable and effective tool in supporting diagnostic decision-making for brain tumor assessment.