Abstract:
Accurate diagnosis of brain tumors using Magnetic Resonance Imaging (MRI) remains challenging due to the labor-intensive and subjective nature of manual segmentation, which introduces variability and increases the risk of misdiagnosis. Conventional automated brain tumor segmentation methods often fail to precisely delineate tumor boundaries, compromising the reliability and precision of segmented outputs (Pandit and Banerjee, 2023). This study evaluates the evolution of deep learning (DL) techniques for MRI analysis, emphasizing their advantages over traditional methodologies in addressing these limitations. DL approaches leverage advanced feature extraction to enhance segmentation accuracy by capturing contextual information derived from image geometry and perspective. To address the critical need for precise tumor boundary detection, we propose a novel dual decoder U-Net architecture integrated with explainable AI (XAI) techniques. This model simultaneously performs edge detection and semantic segmentation, enabling robust identification of tumor boundaries and regions.
