Brain Tumor Classification in MRI Images: A CNN and U-Net Approach

The timely detection of brain tumors is pivotal for improving survival prospects. Employing diagnostic imaging modalities like MRI and CT, this study prioritizes MRI due to its ability to yield intricate images of tissues and organs compared to CT scans. The research employs two distinct methodologies: Convolutional Neural Network (CNN) and U-Net. Initially, images from the Kaggle dataset undergo meticulous segmentation into training, validation, and test datasets, categorizing tumor and non-tumor sections. Subsequently, image processing incorporates a Gaussian filter. Precise segmentation of dataset images follows. Deep learning models, CNN and U-Net, are then trained using the segmented images. Performance metrics, including accuracy, precision, recall, and area under the curve (AUC), are assessed, revealing impressive scores of 94, 94, 98, and 99% for the CNN model. In conclusion, the models are applied for tumor predictions, with the CNN model outperforming the U-Net model. This research not only advances early brain tumor diagnosis but also underscores CNN’s effectiveness in achieving optimal performance metrics.