Classification of brain tumor with MR images with a new deep learning-based approach

Accurate and rapid diagnosis of brain tumors is of critical importance to increase the quality of life of patients and improve treatment processes. However, manual examination of MRI images by doctors is a time-consuming and error-prone process. This study aims to automatically classify brain tumors by developing an artificial intelligence-based deep learning model. Within the scope of the study, a dataset consisting of MRI images obtained from open access platforms was used and this dataset was optimized with preprocessing techniques. This dataset was integrated into the study by dividing the proposed model into training, validation and test groups. The model is based on the convolutional neural network (CNN) structure and uses the L1-Norm Support Vector Machines (SVM) ReliefF algorithm for feature selection. In this way, both computational costs are reduced and classification accuracy is increased. Model performance was evaluated with various metrics such as accuracy, sensitivity, specificity, F1 score and ROC curve. This model, which reached 95% accuracy rate especially in the classification of glioblastoma multiforme tumor, provided a significant improvement compared to existing approaches. The results obtained show that the developed model makes a significant contribution to producing automatic decision support systems in the healthcare sector.