Medical image segmentation remains an open research problem due to the inherent complexity of anatomical structures and the variability in imaging modalities with the growth of large clinical datasets, segmentation has become essential for accurate diagnosis and treatment planning in biomedical imaging. Recent advances in deep learning have furthered segmentation techniques, with architectures like the proposed Recurrent Dilated U-Net showing promising results. This model integrates U-Net’s skip connections, recurrent blocks for sequence modelling, and dilated convolutions for multi-scale feature extraction, enhancing segmentation performance. Validated on the ISIC 2018 and Kvasir-SEG datasets, the Recurrent Dilated U-Net achieved high mean Intersection over Union (mIoU) scores of 0.8541 and 0.8782, respectively, underscoring its effectiveness in medical image analysis.