Laser & Optoelectronics Progress, Volume. 62, Issue 2, 0217003(2025)
Retinal Vessel Segmentation Using Multi-Directional Stripe Convolution and Pyramid Dual Pooling
Figures & Tables(13)
Fig. 9. Visual comparison of ablation experiments. (a1) (a2) Images from CHASE-DB1 test set; (b1) (b2) images from DRIVE test set
Table 1. Comparison of results by different segmentation methods
View tableTable 1. Comparison of results by different segmentation methods
Method Year CHASE-DB1 dataset DRIVE dsataset AUC F1 Acc AUC F1 Acc Ref. ［ 23 ］2018 0.9839 0.8031 0.9656 0.9793 0.8202 0.9561 Ref. ［ 24 ］2018 0.9825 — 0.9637 0.9807 — 0.9567 Ref. ［ 12 ］2019 0.9666 0.8006 0.9703 0.9714 0.8091 0.9630 Ref. ［ 25 ］2019 0.9812 0.8037 0.9661 0.9772 0.8270 0.9567 Ref. ［ 14 ］2020 — — — 0.9794 0.8194 0.9561 Ref. ［ 26 ］2020 0.9832 — 0.9694 0.9840 — 0.9610 Ref. ［ 27 ］2021 — 0.7893 0.9711 — 0.8131 0.9660 Ref. ［ 28 ］2022 — 0.7520 0.9660 — 0.8080 0.9660 Ref. ［ 29 ］2022 0.9818 0.7741 0.9714 0.9828 0.8149 0.9678 Ref. ［ 30 ］2022 0.9858 — 0.9732 0.9832 — 0.9675 Ref. ［ 31 ］2023 0.9080 0.7896 0.9708 0.8975 0.8101 0.9650 Ref. ［ 32 ］2023 — 0.8031 0.9612 — 0.8129 0.9589 Ref. ［ 15 ］2023 — 0.7992 0.9737 — 0.8155 0.9668 MPU-Net 2024 0.9863 0.8046 0.9746 0.9851 0.8162 0.9689
Table 2. Performance comparison between MPU-Net and other models
View tableTable 2. Performance comparison between MPU-Net and other models
Model AUC F1 Acc UNet［ 8 ］0.9739 0.8079 0.9673 UNet++［ 9 ］0.9778 0.8136 0.9682 ResUnet［ 33 ］0.9780 0.8118 0.9680 ACC-UNet［ 34 ］0.9797 0.8111 0.9681 MPU-Net （proposed） 0.9851 0.8162 0.9689
Table 3. Results of ablation experiments on CHASE-DB1 dataset
View tableTable 3. Results of ablation experiments on CHASE-DB1 dataset
Model AUC F1 Acc U-Net 0.9813 0.7870 0.9718 U-Net+FDSC 0.9842 0.8039 0.9735 U-Net+PDPF2 0.9809 0.7883 0.9730 U-Net+CSDA 0.9807 0.7887 0.9723 MPU-Net 0.9863 0.8046 0.9746
Table 4. Results of ablation experiments on DRIVE dataset
View tableTable 4. Results of ablation experiments on DRIVE dataset
Model AUC F1 Acc U-Net 0.9739 0.8079 0.9673 U-Net+FDSC 0.9824 0.8160 0.9681 U-Net+PDPF2 0.9742 0.8089 0.9673 U-Net+CSDA 0.9747 0.8087 0.9676 MPU-Net 0.9851 0.8162 0.9689
Tools
Get Citation
Copy Citation Text
Linfeng Kong, Yun Wu. Retinal Vessel Segmentation Using Multi-Directional Stripe Convolution and Pyramid Dual Pooling[J]. Laser & Optoelectronics Progress, 2025, 62(2): 0217003
Paper Information
Category: Medical Optics and Biotechnology
Received: Apr. 15, 2024
Accepted: May. 24, 2024
Published Online: Dec. 17, 2024
The Author Email:
CSTR:32186.14.LOP241105
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20