Brain tumor segmentation and classification Using mrimages

Abstract

Brain tumors impact the brain and spinal cord, making them one of the most severe cancers. Many computer vision techniques have been proposed to aid early diagnosis and reduce surgical intervention. However, these approaches struggle with segmentation and classification of brain tumors in magnetic resonance imaging (MRIs). A new automated brain tumor detection and classification method is proposed in this research. The system includes augmentation, segmentation, and classification. These parts finish the system. An MRI picture is corrected during the enhancing phase using Adaptive Histogram Equalization (AHE). U-NET was needed to distinguish aberrant cells from healthy brain tissue to complete segmentation. The brain tumor's HGG or LGG status was determined using 3D-CNN. To validate the Brats-2015 dataset-based system, many tests were run. The system achieved segmentation accuracy rates of 97% and 99% using 5-fold and 10-fold methods, respectively, resulting in a Dice Similarity Coefficient (DSC) accuracy rate of 99.8%.