MRI Based Brain Tumor Segmentation with Enhanced U-Net Deep Learning Model

Manually identifying brain tumors from MRI scans is challenging and time-consuming, requiring significant effort and increasing the likelihood of errors. Accurate and consistent brain tumor segmentation is essential for cancer detection, treatment planning, and outcome assessment. Effective tumor segmentation is crucial for creating individualized treatment plans, as the early identification of cerebral neoplasms is critical to optimizing therapeutic success. However, identifying brain tumors from MRI images remains a complex and demanding task due to variability in tumor size, shape, and location. To overcome these limitations, this study proposes an Optimized Attention-Driven U-Net with Pyramid Vision Transformers (OA-U-Net-PVT) for enhanced brain tumor segmentation. This model integrates attention mechanisms and pyramid vision transformer modules into the U-Net framework, improving contextual understanding and spatial precision. It achieves a Dice coefficient of 88.7 % and an overall accuracy of 98.7 %, outperforming existing state-of-the-art models. The OA-U-Net-PVT shows strong potential for accurate, automated, and real-time brain tumor detection in clinical applications.