AUTOMATIC BIOMEDICAL BRAIN TUMOR SEGMENTATION FOR DISEASE DIAGNOSIS USING ARTIFICIAL INTELLIGENCE

Brain Tumor problems can lead to various health issues like Nausea, Memory Loss and vision problems andhave a significant impact on daily life which requires early diagnosis and treatment to get saved from furtherdamage and complications. Recent enhancement in Image processing and Artificial Intelligence have broughtnotable transformations in healthcare technology, leading to significant improvements in diagnosis accuracy,cost-effectiveness, and time efficiency. Magnetic Resonance Imaging (MRI) is employed by the radiologist forits remarkable ability to detect even the most critical and tiny brain abnormalities. This study considers acomprehensive segmentation frameworks – U-Net, U-Net++ and Modified U-Net, leveraging state of-the-artdeep learning architectures to segment brain cancers within MRI. To confirm detailed segmentation of the tumorregions, the preprocessing phase includes advanced mask alignment techniques. This pre-processed dataset hasbeen used to evaluate the performance of the deep learning models for brain tumor segmentation. Modified UNet backbone architecture-based customized deep-learning model is built to segment brain tumors whichenhances feature extraction and multi-scale feature fusion. From the numerical results of modified U-Net, it wasnoticed that a validation accuracy of 0.99756, validation IoU of 0.8014 and dice co efficient of 0.8880. Incomparison, modified U Net exhibits strong performance with box detection accuracy of 0.99768, validationIoU of 0.8131 and dice co efficient of 0.8952. To underscore the novelty of the approach, the performance of theproposed framework is systematically compared with established methods such as U-Net, U-Net. These findingsshowcase the robustness, generalizability and effectiveness of the proposed modified framework for properimage segmentation contributing towards advanced medical technologies and diagnosis precision..