DCRU-Net: Dynamic Contextual Residual U-Net for Medical Image Segmentation
DOI:
https://doi.org/10.47852/bonviewJDSIS52024977Keywords:
deep learning, medical image segmentation, DCRU-Net, computer-aided diagnosisAbstract
Accurate medical image segmentation (MIS) is crucial for computer-assisted diagnosis and treatment planning. This research proposes a deep learning (DL) architecture for accurate and efficient MIS, named Dynamic Contextual Residual U-Net (DCRU-Net). This design is a variation of the conventional U-Net that combines dynamic contextual residual block (DCRB) with a squeeze-and-excitation (SE) block. DCRU-Net combines the strengths of the DCRB and SE block. The SE block improves the feature-capture performance of the model by retuning the channel-specific feature responses. The DCRB adaptively modifies feature representations, selecting and adding significantly relevant contextual features at each network step, making DCRU-Net adaptable across multiple medical imaging modalities and to the difficulties of segmentation tasks. Experimental tests on six medical image collections show that DCRU-Net is superior to state-of-the-art (SOTA) methods in terms of Dice similarity coefficients (DSC) and intersection over union (IoU). Consistent performance is achieved across different medical imaging datasets with less annotated data because of the resilience and generalizability of the architecture. DCRU-Net is a new approach to accurate and automated MIS that can transform healthcare by improving segmentation accuracy and flexibility and becoming an invaluable instrument in computer-aided diagnosis.
Received: 9 December 2024 | Revised: 20 January 2025 | Accepted: 4 July 2025
Conflicts of Interest
The authors declare that they have no conflicts of interest to this work.
Data Availability Statement
The data that support the findings of this study are openly available in Kaggle at https://www.kaggle.com/competitions/data-sci-ence-bowl-2018. The data that support the findings of this study are openly available in Google Drive at https://drive.google.com/drive/folders/10QXjxBJqCf7PAXqbDvoceWmZ-qF07tFi?usp=share_link. The data that support the findings of this study are openly available in Polyp DataSet at https://doi.org/10.6084/m9.figshare.21221579.v2.
Author Contribution Statement
Manoj Kumar Singh: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review and editing, Visualization. Satish Chand: Validation, Formal analysis, Writing – review and editing, Visualization. Devender Kumar: Formal analysis, Writing – review and editing, Visualization, Supervision, Project administration.
Downloads
Published
Issue
Section
License
Copyright (c) 2025 Authors

This work is licensed under a Creative Commons Attribution 4.0 International License.