Hybrid Deep Learning and Reinforcement Learning Approach for Brain Tumor Classification from MRI Images

---

Saraç Ç. G., ARIBAL S. A., Üncü Y. A.

Süleyman Demirel Üniversitesi Fen Edebiyat Fakültesi Fen Dergisi, cilt.20, sa.2, ss.206-221, 2025 (TRDizin)

• Yayın Türü: Makale / Tam Makale
• Cilt numarası: 20 Sayı: 2
• Basım Tarihi: 2025
• Doi Numarası: 10.29233/sdufeffd.1694369
• Dergi Adı: Süleyman Demirel Üniversitesi Fen Edebiyat Fakültesi Fen Dergisi
• Derginin Tarandığı İndeksler: zbMATH, Directory of Open Access Journals, TR DİZİN (ULAKBİM)
• Sayfa Sayıları: ss.206-221
• Akdeniz Üniversitesi Adresli: Evet

Özet

Brain cancer, resulting from abnormal tumor growth in brain tissue, requires accurate and timely diagnosis. Although MRI plays a crucial role, manual interpretation is prone to errors and delays. To address this, we propose a hybrid system combining deep learning (VGG16, ResNet50, DenseNet201) with reinforcement learning (Q-learning) for brain tumor classification. Using three distinct MRI datasets within MATLAB, the models achieved high classification accuracies: 97.45% (VGG16), 96.06% (ResNet50), and 96.93% (DenseNet201). The integration of reinforcement learning improved decision-making and interpretability. Additionally, a user-friendly interface was developed to support clinical decision-making. This study demonstrates that combining deep learning with reinforcement learning enhances model adaptability, offering a more reliable and effective diagnostic approach.