Deep Learning Approach to Classify Brain Tumors from Magnetic Resonance Imaging Images

Brain tumor is one of the primary causes of mortality all over the globe, and it poses as one of the most complicated tasks in contemporary medicine when it comes to its proper diagnosis and classification into its many different types. Both benign and malignant tumors affect the lives of their respective patients as they may lead to mortality, or in the least many related difficulties and sicknesses. Typically, MRI (Magnetic Resonance Imaging) is used as a diagnostic technique where experts manually analyze the images to detect tumors. On the other hand, advanced technologies such as deep learning can step into the light and aid in the diagnosis and classification procedures in a much more time-efficient and precise manner. MRI images are an effective input that can be used in deep learning technologies such as CNN in order to accurately detect brain tumors. In this study, VGG-16, ResNet50, and Xception were trained on a Kaggle dataset consisting of brain tumor MRI images. The performance of the models was evaluated where it was found that brain tumors can be efficiently detected from MRI images with high accuracy and precision using VGG-16, ResNet50, and Xception. The highest performing model was the proposed XCeption model with perfect scores.