Development of interface for the diagnosis and classification of brain cancers with reinforcement learning algorithm supported by deep learning application

Brain cancer is a severe condition resulting from the uncontrolled proliferation of tumors within brain tissue. It manifests through various symptoms, including headaches, nausea, vomiting, balance issues, and neurological impairments. Treatment approaches vary based on tumor type and the patient's overall health and may involve surgery, radiotherapy, chemotherapy, or targeted drug therapies. Magnetic Resonance Imaging (MRI) is a crucial tool in detecting and diagnosing brain tumors. MRI provides high- resolution images, making it possible to detect and evaluate tumors. Medical image processing is a set of digital techniques used to analyze and interpret medical images. In recent years, methods such as artificial intelligence, deep learning and reinforcement learning have made great progress in these processes. In the literature, there have been no research where deep learning architectures and reinforcement learning algorithms are used together. The primary objective of this study is to design an automated diagnostic system for classifying primary brain tumors from MRI images by integrating deep learning and reinforcement learning techniques. To achieve this, a classification system utilizing a Q-learning algorithm in combination with deep learning architectures was developed. Within the Matlab environment, MR images of primary brain tumors were analyzed using three distinct datasets. The proposed model, incorporating the Q-learning algorithm, achieved an accuracy of 97.45% with the VGG16 architecture, 96.06% with ResNet50, and 96.93% with DenseNet201. The findings indicate that the proposed model demonstrates strong potential for brain tumor classification. In addition, the system was supported with an interface for a faster and more reliable diagnostic process for physicians.