Deep learning has revolutionized medical image analysis by enabling rapid, automated interpretation of complex imaging data. We propose a novel framework that combines a self‑configuring U‑Net backbone with integrated transformer‑style attention modules to enhance both local feature extraction and global contextual reasoning. Using publicly available datasets—the Synapse multi‑organ CT challenge and the ACDC cardiac MRI collection—we demonstrate that our method achieves mean Dice scores of 0.83 and 0.77, respectively, with competitive 95th‑percentile Hausdorff distances, all under CPU‑only training conditions. Qualitative visualizations, per‑class performance analysis, and volume‑versus‑accuracy studies reveal robust segmentation of challenging anatomical structures. Comparative evaluation against CE‑Net, TransUNet, and 3D TransUNet shows that our approach approaches state‑of‑the‑art benchmarks without extensive manual tuning. Our framework’s self‑configuration and attention mechanisms offer a reproducible, resource‑efficient solution for clinical deployment and pave the way for further enhancements through multi‑modal fusion and domain adaptation
Afaf E.G. Osman (Wed,) studied this question.