<p>Image segmentation is critical in computer vision, particularly in partitioning images into meaningful segments or regions for analysis. Conventional methods often face challenges in handling complex image structures and generalizing across diverse datasets. To address these limitations, this paper introduces AGU-Net, a novel attention-guided U-Net architecture designed specifically for medical image segmentation. AGU-Net leverages attention mechanisms to focus on informative regions of the image, improving feature representation and localization accuracy. Our key contributions include the integration of a Dual Attention Focus (Spatial and Channel Attention) attention-guided mechanism into U-Net, enhancing its ability to address challenges such as limited annotated data, and the development of an attention-guided loss function that prioritizes critical regions during training, further improving segmentation performance and U-Net generalizability. Experimental results in multiple medical imaging datasets demonstrate that AGU-Net outperforms conventional convolutional neural networks, achieving improvements in segmentation precision ranging from 5% on standard datasets to 28% when data augmentation is applied to improve dataset variability. These contributions collectively advance the state of the art in medical image segmentation, offering a robust and interpretable solution for clinical applications.</p>