AI-Driven Approaches for Brain Tumor Segmentation

Brain tumor segmentation is a critical step in medical image analysis, playing a crucial role in diagnosis, treatment planning, and disease monitoring. Accurate and efficient segmentation enables early detection and personalized therapeutic strategies. However, traditional manual segmentation is time-consuming, making automated approaches highly desirable. Recent deep learning techniques, particularly convolutional neural networks (CNNs), have significantly improved segmentation performance. This study investigates and compares the performance of two deep learning-based segmentation models, U-Net and Mask R-CNN, applied to FLAIR MRI (Magnetic Resonance Imaging) images from 484 patients with brain tumors, based on the publicly available BRATS (Brain Tumor Segmentation Challenge) dataset. Both models were trained to generate masks highlighting tumor regions. U-Net outperformed Mask R-CNN in overall segmentation accuracy and inference speed, making it suitable for fast clinical assessments. In contrast, Mask R-CNN achieved more precise boundary delineation, which could be advantageous in treatment planning or surgical navigation. The findings suggest that each model may be better suited for specific clinical tasks and emphasize the importance of model selection in AI-assisted neuroimaging. Our findings highlight the strengths and limitations of each approach, providing insights into their clinical applicability and contributing to efforts toward integrating AI into medical imaging workflows to enhance diagnostic efficiency and patient outcomes.