IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting
Abstract In clinical medicine, the segmentation of blood vessels in retinal images is essential for subsequent analysis in clinical diagnosis. However, retinal images are often noisy and their vascular structure is relatively tiny, which poses significant challenges for vessel segmentation. To impro...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2022-11-01
|
| Series: | IET Image Processing |
| Online Access: | https://doi.org/10.1049/ipr2.12580 |
| _version_ | 1817980673201274880 |
|---|---|
| author | M. Zhu K. Zeng G. Lin Y. Gong T. Hao K. Wattanachote X. Luo |
| author_facet | M. Zhu K. Zeng G. Lin Y. Gong T. Hao K. Wattanachote X. Luo |
| author_sort | M. Zhu |
| collection | DOAJ |
| description | Abstract In clinical medicine, the segmentation of blood vessels in retinal images is essential for subsequent analysis in clinical diagnosis. However, retinal images are often noisy and their vascular structure is relatively tiny, which poses significant challenges for vessel segmentation. To improve the performance of vessel segmentation, an improved model IterNet++ based on the architecture of IterNet is proposed. First, curvelet signal analysis is applied to enhance retinal images. Second, residual convolution (ResConv) blocks and guided filters are introduced to utilise the encoder features of previous iterations in the model to reduce overfitting. Third, offline hard‐sample mining is used to improve segmentation performance by utilising training samples with low segmentation accuracy as many possible on a few‐sample training set. In addition, a test‐time augmentation method is applied to testing samples in test dataset during inference. Extensive experiments show that this model achieves Dice scores of 0.8313, 0.8277, and 0.8372 on DRIVE, CHASE‐DB1, and STARE datasets, respectively, demonstrating the best performance compared with IterNet and other baseline models. |
| first_indexed | 2024-04-13T22:56:22Z |
| format | Article |
| id | doaj.art-075140ba29fc46c9b3d66f4da8805e97 |
| institution | Directory Open Access Journal |
| issn | 1751-9659 1751-9667 |
| language | English |
| last_indexed | 2024-04-13T22:56:22Z |
| publishDate | 2022-11-01 |
| publisher | Wiley |
| record_format | Article |
| series | IET Image Processing |
| spelling | doaj.art-075140ba29fc46c9b3d66f4da8805e972022-12-22T02:25:59ZengWileyIET Image Processing1751-96591751-96672022-11-0116133617363310.1049/ipr2.12580IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmentingM. Zhu0K. Zeng1G. Lin2Y. Gong3T. Hao4K. Wattanachote5X. Luo6School of Computer Science and Engineering Sun Yat‐Sen University Guangzhou ChinaSchool of Computer Science and Engineering Sun Yat‐Sen University Guangzhou ChinaSchool of Computer Science and Engineering Sun Yat‐Sen University Guangzhou ChinaGuangdong University of Foreign Studies Guangzhou ChinaSchool of Computer Science South China Normal University Guangzhou ChinaGuangdong University of Foreign Studies Guangzhou ChinaNational & Local Joint Engineering Research Center of Satellite Navigation and Location Service Guilin University of Electronic Technology Guilin ChinaAbstract In clinical medicine, the segmentation of blood vessels in retinal images is essential for subsequent analysis in clinical diagnosis. However, retinal images are often noisy and their vascular structure is relatively tiny, which poses significant challenges for vessel segmentation. To improve the performance of vessel segmentation, an improved model IterNet++ based on the architecture of IterNet is proposed. First, curvelet signal analysis is applied to enhance retinal images. Second, residual convolution (ResConv) blocks and guided filters are introduced to utilise the encoder features of previous iterations in the model to reduce overfitting. Third, offline hard‐sample mining is used to improve segmentation performance by utilising training samples with low segmentation accuracy as many possible on a few‐sample training set. In addition, a test‐time augmentation method is applied to testing samples in test dataset during inference. Extensive experiments show that this model achieves Dice scores of 0.8313, 0.8277, and 0.8372 on DRIVE, CHASE‐DB1, and STARE datasets, respectively, demonstrating the best performance compared with IterNet and other baseline models.https://doi.org/10.1049/ipr2.12580 |
| spellingShingle | M. Zhu K. Zeng G. Lin Y. Gong T. Hao K. Wattanachote X. Luo IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting IET Image Processing |
| title | IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting |
| title_full | IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting |
| title_fullStr | IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting |
| title_full_unstemmed | IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting |
| title_short | IterNet++: An improved model for retinal image segmentation by curvelet enhancing, guided filtering, offline hard‐sample mining, and test‐time augmenting |
| title_sort | iternet an improved model for retinal image segmentation by curvelet enhancing guided filtering offline hard sample mining and test time augmenting |
| url | https://doi.org/10.1049/ipr2.12580 |
| work_keys_str_mv | AT mzhu iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT kzeng iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT glin iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT ygong iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT thao iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT kwattanachote iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting AT xluo iternetanimprovedmodelforretinalimagesegmentationbycurveletenhancingguidedfilteringofflinehardsampleminingandtesttimeaugmenting |