Multi-class segmentation of knee MRI based on hybrid attention
IntroductionAccurate segmentation of knee MRI images is crucial for the diagnosis and treatment of degenerative knee disease and sports injuries. However, many existing methods are hindered by class imbalance and fail to capture the features of small structures, leading to suboptimal segmentation pe...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-06-01
|
| Series: | Frontiers in Medicine |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fmed.2025.1581487/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849726876583460864 |
|---|---|
| author | Yuhang Xiang Xinglin Zhang Tao Meng Tao Meng Tao Chen Tao Chen |
| author_facet | Yuhang Xiang Xinglin Zhang Tao Meng Tao Meng Tao Chen Tao Chen |
| author_sort | Yuhang Xiang |
| collection | DOAJ |
| description | IntroductionAccurate segmentation of knee MRI images is crucial for the diagnosis and treatment of degenerative knee disease and sports injuries. However, many existing methods are hindered by class imbalance and fail to capture the features of small structures, leading to suboptimal segmentation performance.MethodsThis study applies hybrid attention and multi-scale feature extraction methods to the problem of multi-class segmentation of knee MRI images and innovates the classic U-Net architecture. Firstly, we propose a Hierarchical Feature Enhancement Fusion (HFEF) module, which is integrated into both the skip connections and the bottleneck layer. This module captures channel and spatial information at multiple levels, enabling the model to efficiently combine local and global features. Secondly, we introduce the Atrous Squeeze Attention (ASA) module, which enables the model to focus on multi-scale features and capture long-range dependencies, thereby improving the segmentation accuracy of complex multi-class structures. Lastly, the loss function is optimized to address the challenges of class imbalance and limited data. The improved loss function enhances the model's ability to learn underrepresented classes, thus enhancing the overall segmentation performance.ResultsWe evaluated the proposed method on a knee MRI dataset and compared it with U-Net. HASA-ResUNet achieved a 12.12% improvement in Intersection over Union (IoU) for the low-frequency and small-sized class, the anterior cruciate ligament, and a 3.32% improvement in mean Intersection over Union (mIoU) across all classes.ConclusionThese results demonstrate that the proposed hybrid attention and multi-scale strategy can effectively address the challenges of class imbalance in knee MRI images, improving the model's overall segmentation performance. |
| format | Article |
| id | doaj-art-0fc22da9d0b149cbb70ca9eeba6433f3 |
| institution | DOAJ |
| issn | 2296-858X |
| language | English |
| publishDate | 2025-06-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Medicine |
| spelling | doaj-art-0fc22da9d0b149cbb70ca9eeba6433f32025-08-20T03:10:02ZengFrontiers Media S.A.Frontiers in Medicine2296-858X2025-06-011210.3389/fmed.2025.15814871581487Multi-class segmentation of knee MRI based on hybrid attentionYuhang Xiang0Xinglin Zhang1Tao Meng2Tao Meng3Tao Chen4Tao Chen5School of Medical Information Engineering, Gannan Medical University, Ganzhou, ChinaShanghai Medical Image Insights Intelligent Technology Co., Ltd., Shanghai, ChinaShanghai Medical Image Insights Intelligent Technology Co., Ltd., Shanghai, ChinaJiangxi Rimag Group Co., Ltd., Nanchang, ChinaBig Data Research Lab, University of Waterloo, Waterloo, ON, CanadaLabor and Worklife Program, Harvard University, Cambridge, MA, United StatesIntroductionAccurate segmentation of knee MRI images is crucial for the diagnosis and treatment of degenerative knee disease and sports injuries. However, many existing methods are hindered by class imbalance and fail to capture the features of small structures, leading to suboptimal segmentation performance.MethodsThis study applies hybrid attention and multi-scale feature extraction methods to the problem of multi-class segmentation of knee MRI images and innovates the classic U-Net architecture. Firstly, we propose a Hierarchical Feature Enhancement Fusion (HFEF) module, which is integrated into both the skip connections and the bottleneck layer. This module captures channel and spatial information at multiple levels, enabling the model to efficiently combine local and global features. Secondly, we introduce the Atrous Squeeze Attention (ASA) module, which enables the model to focus on multi-scale features and capture long-range dependencies, thereby improving the segmentation accuracy of complex multi-class structures. Lastly, the loss function is optimized to address the challenges of class imbalance and limited data. The improved loss function enhances the model's ability to learn underrepresented classes, thus enhancing the overall segmentation performance.ResultsWe evaluated the proposed method on a knee MRI dataset and compared it with U-Net. HASA-ResUNet achieved a 12.12% improvement in Intersection over Union (IoU) for the low-frequency and small-sized class, the anterior cruciate ligament, and a 3.32% improvement in mean Intersection over Union (mIoU) across all classes.ConclusionThese results demonstrate that the proposed hybrid attention and multi-scale strategy can effectively address the challenges of class imbalance in knee MRI images, improving the model's overall segmentation performance.https://www.frontiersin.org/articles/10.3389/fmed.2025.1581487/fullmedical image segmentationdeep learningattention mechanismkneeMRI |
| spellingShingle | Yuhang Xiang Xinglin Zhang Tao Meng Tao Meng Tao Chen Tao Chen Multi-class segmentation of knee MRI based on hybrid attention Frontiers in Medicine medical image segmentation deep learning attention mechanism knee MRI |
| title | Multi-class segmentation of knee MRI based on hybrid attention |
| title_full | Multi-class segmentation of knee MRI based on hybrid attention |
| title_fullStr | Multi-class segmentation of knee MRI based on hybrid attention |
| title_full_unstemmed | Multi-class segmentation of knee MRI based on hybrid attention |
| title_short | Multi-class segmentation of knee MRI based on hybrid attention |
| title_sort | multi class segmentation of knee mri based on hybrid attention |
| topic | medical image segmentation deep learning attention mechanism knee MRI |
| url | https://www.frontiersin.org/articles/10.3389/fmed.2025.1581487/full |
| work_keys_str_mv | AT yuhangxiang multiclasssegmentationofkneemribasedonhybridattention AT xinglinzhang multiclasssegmentationofkneemribasedonhybridattention AT taomeng multiclasssegmentationofkneemribasedonhybridattention AT taomeng multiclasssegmentationofkneemribasedonhybridattention AT taochen multiclasssegmentationofkneemribasedonhybridattention AT taochen multiclasssegmentationofkneemribasedonhybridattention |