A CNN Approach to Automated Detection and Classification of Brain Tumors

For a timely diagnosis and successful treatment, brain tumors need to be carefully evaluated. Tumor inspection is complicated by morphological factors like size, location, texture, and varying appearance. Additionally, medical imaging poses challenges such as noise and incomplete data. This research article presents a methodology for processing Magnetic Resonance Imaging (MRI) data, encompassing techniques for image classification and denoising. The effective use of MRI images allows medical professionals to detect brain disorders, including tumors. This study uses MRI data analysis to classify brain tumors and healthy brain tissue. MRI is the best imaging technique for studying brain tumors because it offers a more thorough image of internal anatomical features than other imaging techniques like Computed Tomography (CT). Initially, an anisotropic diffusion filter is used to denoise the MRI picture. The models are based on an open-access, clinically validated MRI dataset for brain tumor classification, containing 3,264 brain MRI scans. SMOTE (Synthetic Minority Over-sampling Technique) is a widely used approach for tackling class imbalance in datasets by generating synthetic samples to augment and balance the data. It is utilized for dataset augmentation and balancing. Convolutional Neural Networks (CNNs), including ResNet152V2, VGG, ViT, and EfficientNet, were utilized for the classification task. Among them, EfficientNet attained an accuracy of 98%, the highest recorded.