CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion
Retinal vessel segmentation plays a critical role in the diagnosis and treatment of various ophthalmic diseases. However, due to poor image contrast, intricate vascular structures, and limited datasets, retinal vessel segmentation remains a long-term challenge. In this paper, based on an encoder–dec...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-11-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/12/23/4743 |
| _version_ | 1797400270310735872 |
|---|---|
| author | Yanan Gu Ruyi Cao Dong Wang Bibo Lu |
| author_facet | Yanan Gu Ruyi Cao Dong Wang Bibo Lu |
| author_sort | Yanan Gu |
| collection | DOAJ |
| description | Retinal vessel segmentation plays a critical role in the diagnosis and treatment of various ophthalmic diseases. However, due to poor image contrast, intricate vascular structures, and limited datasets, retinal vessel segmentation remains a long-term challenge. In this paper, based on an encoder–decoder framework, a novel retinal vessel segmentation model called CMP-UNet is proposed. Firstly, the Coarse and Fine Feature Aggregation module decouples and aggregates coarse and fine vessel features using two parallel branches, thus enhancing the model’s ability to extract features for vessels of various sizes. Then, the Multi-Scale Channel Adaptive Fusion module is embedded in the decoder to realize the efficient fusion of cascade features by mining the multi-scale context information from these features. Finally, to obtain more discriminative vascular features and enhance the connectivity of vascular structures, the Pyramid Feature Fusion module is proposed to effectively utilize the complementary information of multi-level features. To validate the effectiveness of the proposed model, it is evaluated on three publicly available retinal vessel segmentation datasets: CHASE_DB1, DRIVE, and STARE. The proposed model, CMP-UNet, reaches F1-scores of 82.84%, 82.55%, and 84.14% on these three datasets, with improvements of 0.76%, 0.31%, and 1.49%, respectively, compared with the baseline. The results show that the proposed model achieves higher segmentation accuracy and more robust generalization capability than state-of-the-art methods. |
| first_indexed | 2024-03-09T01:53:08Z |
| format | Article |
| id | doaj.art-a3fa55d3d819428d870b2d1fb1e03ba1 |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-09T01:53:08Z |
| publishDate | 2023-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-a3fa55d3d819428d870b2d1fb1e03ba12023-12-08T15:13:50ZengMDPI AGElectronics2079-92922023-11-011223474310.3390/electronics12234743CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature FusionYanan Gu0Ruyi Cao1Dong Wang2Bibo Lu3School of Computer Science and Technology, Henan Polytechnic University, Jiaozuo 454003, ChinaSchool of Computer Science and Technology, Henan Polytechnic University, Jiaozuo 454003, ChinaSchool of Mathematics/S.T. Yau Center of Southeast University, Southeast University, Nanjing 210096, ChinaSchool of Computer Science and Technology, Henan Polytechnic University, Jiaozuo 454003, ChinaRetinal vessel segmentation plays a critical role in the diagnosis and treatment of various ophthalmic diseases. However, due to poor image contrast, intricate vascular structures, and limited datasets, retinal vessel segmentation remains a long-term challenge. In this paper, based on an encoder–decoder framework, a novel retinal vessel segmentation model called CMP-UNet is proposed. Firstly, the Coarse and Fine Feature Aggregation module decouples and aggregates coarse and fine vessel features using two parallel branches, thus enhancing the model’s ability to extract features for vessels of various sizes. Then, the Multi-Scale Channel Adaptive Fusion module is embedded in the decoder to realize the efficient fusion of cascade features by mining the multi-scale context information from these features. Finally, to obtain more discriminative vascular features and enhance the connectivity of vascular structures, the Pyramid Feature Fusion module is proposed to effectively utilize the complementary information of multi-level features. To validate the effectiveness of the proposed model, it is evaluated on three publicly available retinal vessel segmentation datasets: CHASE_DB1, DRIVE, and STARE. The proposed model, CMP-UNet, reaches F1-scores of 82.84%, 82.55%, and 84.14% on these three datasets, with improvements of 0.76%, 0.31%, and 1.49%, respectively, compared with the baseline. The results show that the proposed model achieves higher segmentation accuracy and more robust generalization capability than state-of-the-art methods.https://www.mdpi.com/2079-9292/12/23/4743retinal vessel segmentationdeep learningmulti-scale feature fusionchannel attention |
| spellingShingle | Yanan Gu Ruyi Cao Dong Wang Bibo Lu CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion Electronics retinal vessel segmentation deep learning multi-scale feature fusion channel attention |
| title | CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion |
| title_full | CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion |
| title_fullStr | CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion |
| title_full_unstemmed | CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion |
| title_short | CMP-UNet: A Retinal Vessel Segmentation Network Based on Multi-Scale Feature Fusion |
| title_sort | cmp unet a retinal vessel segmentation network based on multi scale feature fusion |
| topic | retinal vessel segmentation deep learning multi-scale feature fusion channel attention |
| url | https://www.mdpi.com/2079-9292/12/23/4743 |
| work_keys_str_mv | AT yanangu cmpunetaretinalvesselsegmentationnetworkbasedonmultiscalefeaturefusion AT ruyicao cmpunetaretinalvesselsegmentationnetworkbasedonmultiscalefeaturefusion AT dongwang cmpunetaretinalvesselsegmentationnetworkbasedonmultiscalefeaturefusion AT bibolu cmpunetaretinalvesselsegmentationnetworkbasedonmultiscalefeaturefusion |