SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction
With the rapid development of convolutional neural networks in image processing, deep learning has been widely applied to medical image segmentation tasks, including liver, retinal vessels, nuclei, and COVID-19 lesion segmentation. However, the accuracy and interpretability of these methods still ne...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10829919/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841542519185211392 |
|---|---|
| author | Yinuo Cao Yong Cheng |
| author_facet | Yinuo Cao Yong Cheng |
| author_sort | Yinuo Cao |
| collection | DOAJ |
| description | With the rapid development of convolutional neural networks in image processing, deep learning has been widely applied to medical image segmentation tasks, including liver, retinal vessels, nuclei, and COVID-19 lesion segmentation. However, the accuracy and interpretability of these methods still need further improvement. This paper proposes a novel Shape-aware U-Net architecture with Attention Mechanism and Context Extraction (SACU-Net) to address the aforementioned issues regarding segmentation performance and shape identification. Our model introduces three major modifications to the standard U-Net: 1) a new context extraction module block (MRC) to capture high-level context features, 2) redesigned skip pathways with spatial and channel attention mechanisms to reduce the semantic gap between the feature maps of encoder and decoder sub-networks, and 3) a novel loss function to enhance segmentation accuracy. We have evaluated SACU-Net against various U-Net variants on four different medical image segmentation tasks: liver segmentation in abdominal CT scans, retinal vessel segmentation, nuclei segmentation, and COVID-19 lesion segmentation. The results demonstrate relative improvements in performance of 7.30%, 5.5%, and 8.40% compared to state-of-the-art U-Net variants. |
| format | Article |
| id | doaj-art-0784adc8358b49caa48d4259425dfbdd |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-0784adc8358b49caa48d4259425dfbdd2025-01-14T00:02:43ZengIEEEIEEE Access2169-35362025-01-01135719573010.1109/ACCESS.2025.352660210829919SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context ExtractionYinuo Cao0https://orcid.org/0009-0006-2415-4372Yong Cheng1https://orcid.org/0000-0002-9057-2007Department of Statistics, The Ohio State University, Columbus, OH, USASchool of Information Science and Technology, Beijing University of Chemical Technology, Beijing, ChinaWith the rapid development of convolutional neural networks in image processing, deep learning has been widely applied to medical image segmentation tasks, including liver, retinal vessels, nuclei, and COVID-19 lesion segmentation. However, the accuracy and interpretability of these methods still need further improvement. This paper proposes a novel Shape-aware U-Net architecture with Attention Mechanism and Context Extraction (SACU-Net) to address the aforementioned issues regarding segmentation performance and shape identification. Our model introduces three major modifications to the standard U-Net: 1) a new context extraction module block (MRC) to capture high-level context features, 2) redesigned skip pathways with spatial and channel attention mechanisms to reduce the semantic gap between the feature maps of encoder and decoder sub-networks, and 3) a novel loss function to enhance segmentation accuracy. We have evaluated SACU-Net against various U-Net variants on four different medical image segmentation tasks: liver segmentation in abdominal CT scans, retinal vessel segmentation, nuclei segmentation, and COVID-19 lesion segmentation. The results demonstrate relative improvements in performance of 7.30%, 5.5%, and 8.40% compared to state-of-the-art U-Net variants.https://ieeexplore.ieee.org/document/10829919/Medical image segmentationU-NetCNNdeep learningattention mechanism |
| spellingShingle | Yinuo Cao Yong Cheng SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction IEEE Access Medical image segmentation U-Net CNN deep learning attention mechanism |
| title | SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction |
| title_full | SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction |
| title_fullStr | SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction |
| title_full_unstemmed | SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction |
| title_short | SACU-Net: Shape-Aware U-Net for Biomedical Image Segmentation With Attention Mechanism and Context Extraction |
| title_sort | sacu net shape aware u net for biomedical image segmentation with attention mechanism and context extraction |
| topic | Medical image segmentation U-Net CNN deep learning attention mechanism |
| url | https://ieeexplore.ieee.org/document/10829919/ |
| work_keys_str_mv | AT yinuocao sacunetshapeawareunetforbiomedicalimagesegmentationwithattentionmechanismandcontextextraction AT yongcheng sacunetshapeawareunetforbiomedicalimagesegmentationwithattentionmechanismandcontextextraction |