Automated Brain Tumor Classification From MRI Images Using Convolutional Neural Networks

Brain tumors are a major health problem and the only way to improve the outcome of the patients is to diagnose them accurately and very quickly. MRI scanning is usually the manual method of interpreting which suffers from being slow and varying between radiologists. In this research, a totally new AI method has been proposed that relies mostly on the Convolutional Neural Network (CNN) techniques for brain tumor classification. The model has been prepared and assessed using the Brain Tumor MRI Dataset, which is publicly available, and it covers four different tumor types: glioma, meningioma, pituitary tumor, and no tumor. As part of data preprocessing, normalization, augmentation, and image resizing were performed in order to increase the model's robustness and generalization. The CNN setup was made to automatically pick out the spatial features and appropriate patterns that are linked to tumor identification thus eliminating the manual feature engineering process.