Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Apr 26;15(5):3344-3365.
doi: 10.1364/BOE.522482. eCollection 2024 May 1.

FRD-Net: a full-resolution dilated convolution network for retinal vessel segmentation

Hua Huang  1 , Zhenhong Shang  1   2 , Chunhui Yu  1
Affiliations

FRD-Net: a full-resolution dilated convolution network for retinal vessel segmentation

Hua Huang et al. Biomed Opt Express. .

Abstract

Accurate and automated retinal vessel segmentation is essential for performing diagnosis and surgical planning of retinal diseases. However, conventional U-shaped networks often suffer from segmentation errors when dealing with fine and low-contrast blood vessels due to the loss of continuous resolution in the encoding stage and the inability to recover the lost information in the decoding stage. To address this issue, this paper introduces an effective full-resolution retinal vessel segmentation network, namely FRD-Net, which consists of two core components: the backbone network and the multi-scale feature fusion module (MFFM). The backbone network achieves horizontal and vertical expansion through the interaction mechanism of multi-resolution dilated convolutions while preserving the complete image resolution. In the backbone network, the effective application of dilated convolutions with varying dilation rates, coupled with the utilization of dilated residual modules for integrating multi-scale feature maps from adjacent stages, facilitates continuous learning of multi-scale features to enhance high-level contextual information. Moreover, MFFM further enhances segmentation by fusing deeper multi-scale features with the original image, facilitating edge detail recovery for accurate vessel segmentation. In tests on multiple classical datasets,compared to state-of-the-art segmentation algorithms, FRD-Net achieves superior performance and generalization with fewer model parameters.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Figures

Fig. 1.
Fig. 1.
Visualization of segmentation results with different down-sampling operations on CE-Net. (a) Original image from the DRIVE dataset; (b) Ground truth; (c) Segmentation result of the original CE-Net; (d) Segmentation result of CE-Net without one round of down-sampling; (e) Segmentation result of CE-Net without two rounds of down-sampling.
Fig. 2.
Fig. 2.
The network architecture of FRD-Net
Fig. 3.
Fig. 3.
The order of the dilated convolutions.
Fig. 4.
Fig. 4.
Backbone network of FRD-Net.
Fig. 5.
Fig. 5.
Dilated Residual Module.
Fig. 6.
Fig. 6.
Internal Structure of the MFFM.
Fig. 7.
Fig. 7.
Image Preprocessing: (a) Original image from the DRIVE dataset, (b) Grayscale processed image, (c) Image after CLAHE processing, (d) Image after Gamma Correction.
Fig. 8.
Fig. 8.
Comparative analysis of local regions in retinal vessel. (a) Original image; (b) Ground truth; (c) FRD-Net prediction; (d) FR-UNet prediction; (e) CS2-UNet prediction; (f) SA-UNet prediction; (g) CE-Net prediction.
Fig. 9.
Fig. 9.
Comparative analysis of local regions in retinal vessel. (a) Original image; (b) Ground truth; (c) FRD-Net prediction; (d) FR-UNet prediction; (e) CS2-UNet prediction; (f) SA-UNet prediction; (g) CE-Net prediction.
Fig. 10.
Fig. 10.
Comparison of Challenging Local Areas in Retinal vessel Segmentation.(a) Original Image; (b) Ground Truth; (c) FRD-Net Prediction; (d) FR-UNet Prediction; (e) CS2-UNet Prediction; (f) SA-UNet Prediction; (g) CE-Net Prediction.
Fig. 11.
Fig. 11.
Comparison of Fine Blood Vessel Segmentation between the Proposed Method and Two Observers. (a) Original Image; (b) First Observer; (c) Second Observer; (d) Our Proposed Method.
Fig. 12.
Fig. 12.
F1 and Sensitivity (Se) Scores on the DRIVE Dataset . The values in parentheses represent the number of parameters (in MB). Larger circles indicate a higher number of parameters.
Fig. 13.
Fig. 13.
Visual Results of Ablation Experiments(a) Original image; (b) Local details of the original image; (c) Ground Truth; (d) Baseline + FR + DCEU + MFFM; (e) Baseline + FR + MFFM; (f) Baseline + FR + DCEU; (g) Baseline + FR; (h) Baseline.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Guo Q., Duffy S. P., Matthews K., et al. , “Microfluidic analysis of red blood cell deformability,” J. Biomech. 47(8), 1767–1776 (2014).10.1016/j.jbiomech.2014.03.038 - DOI - PubMed
    1. Xu W., Yang H., Zhang M., et al. , “Decnet: A dual-stream edge complementary network for retinal vessel segmentation,” in 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), (2021), pp. 1595–1600.
    1. Li R.-Q., Xie X.-L., Zhou X.-H., et al. , “Real-time multi-guidewire endpoint localization in fluoroscopy images,” IEEE Trans. Med. Imaging 40(8), 2002–2014 (2021).10.1109/TMI.2021.3069998 - DOI - PubMed
    1. Abràmoff M. D., Garvin M. K., Sonka M., “Retinal imaging and image analysis,” IEEE Rev. Biomed. Eng. 3, 169–208 (2010).10.1109/RBME.2010.2084567 - DOI - PMC - PubMed
    1. Imani E., Javidi M., Pourreza H.-R., “Improvement of retinal blood vessel detection using morphological component analysis,” Comput. Methods Programs Biomed. 118(3), 263–279 (2015).10.1016/j.cmpb.2015.01.004 - DOI - PubMed

LinkOut - more resources