<p>Morphological diversity and edge ambiguity pose challenges for accurate gland segmentation in medical imaging. We propose the MEDU-Net channel: Full-resolution and contour refinement feedback patch-aware U-Net, a novel network integrating three key components to improve segmentation accuracy. First, the full-resolution multilevel fusion module mitigates information loss in the encoder-decoder structure by full-resolution fusion of deep channels, thereby achieving multi-scale feature preservation. Second, the contour refinement perception module adopts multibranch adaptive contour perception to feature refine complex gland edge delineation. Third, the feature feedback patch awareness module in skip connections expands the awareness sensitivity and supervises the feature reshaping process through feature patches and feedback. After evaluation on GlaS and CRAG datasets, MEDU-Net surpasses the state-of-the-art methods in terms of Dice coefficient and segmentation accuracy, and performs well in handling the complexity of glandular structures. This study provides an effective solution for clinical histopathology image segmentation.</p>