Acta Optica Sinica, Volume. 45, Issue 7, 0717001(2025)
LTDA‐Mamba: Retinal Vessel Segmentation Based on a Hybrid CNN‐Mamba Network
Figures & Tables(8)
![Specific structures of VSS module, Mamba layer, and ResMamba block[15]. (a) VSS module; (b) Mamba layer; (c) ResMamba block](/Images/icon/loading.gif){kind=icon}
Fig. 2. Specific structures of VSS module, Mamba layer, and ResMamba block[15]. (a) VSS module; (b) Mamba layer; (c) ResMamba block
Fig. 5. Experimental results with different state-of-the-art models on DRIVE, CHASE_DB1 and STARE datasets. (a) Images; (b) labels; (c) UNet++; (d) CS2Net; (e) SA-UNet; (f) LIOT; (g) LightM-UNet; (h) LTDA-Mamba
Table 1. Quantitative evaluation with different methods
View tableTable 1. Quantitative evaluation with different methods
Dataset Year Method F1 score Accuracy Sensitivity Specificity DRIVE 2020 UNet++ 0.7967 0.9672 0.7296 0.9900 2021 CS2Net 0.8086 0.9676 0.7808 0.9855 2021 SA-UNet 0.8080 0.9676 0.7780 0.9858 2022 LIOT 0.7915 0.9666 0.7274 0.9898 2024 LightM-UNet 0.7953 0.9672 0.7281 0.9896 2024 Ours 0.8151 0.9689 0.7868 0.9867 CHASE_DB1 2020 UNet++ 0.6851 0.9649 0.5568 0.9949 2021 CS2Net 0.8035 0.9736 0.7862 0.9874 2021 SA-UNet 0.7907 0.9729 0.7479 0.9894 2022 LIOT 0.8002 0.9725 0.8193 0.9832 2024 LightM-UNet 0.8005 0.9711 0.7843 0.9852 2024 Ours 0.8043 0.9741 0.7697 0.9894 STARE 2020 UNet++ 0.6594 0.9597 0.5082 0.9972 2021 CS2Net 0.7910 0.9697 0.7472 0.9882 2021 SA-UNet 0.7743 0.9677 0.7228 0.9880 2022 LIOT 0.7541 0.9677 0.6453 0.9945 2024 LightM-UNet 0.8049 0.9734 0.7699 0.9911 2024 Ours 0.8219 0.9792 0.7488 0.9929
Table 2. FLOPs and Params for different methods
View tableTable 2. FLOPs and Params for different methods
Method UNet++ CS2Net SA-UNet LIOT LightM-UNet Ours FLOPs /G 905 185 136 2505 125 132 Params /M 36.63 8.44 0.48 13.60 0.43 0.47
Table 3. Influence of each module on presented framework
View tableTable 3. Influence of each module on presented framework
Dataset Baseline LIOT DCA F1 score Accuracy Sensitivity Specificity DRIVE √ × × 0.7953 0.9672 0.7281 0.9896 √ √ × 0.8092 0.9667 0.7852 0.9866 √ × √ 0.8121 0.9675 0.8002 0.9839 √ √ √ 0.8151 0.9689 0.7868 0.9867 CHASE_DB1 √ × × 0.8005 0.9711 0.7843 0.9852 √ √ × 0.8011 0.9732 0.8062 0.9849 √ × √ 0.8024 0.9737 0.7731 0.9894 √ √ √ 0.8043 0.9741 0.7697 0.9894 STARE √ × × 0.8049 0.9734 0.7699 0.9911 √ √ × 0.8161 0.9746 0.7711 0.9906 √ × √ 0.8152 0.9740 0.7705 0.9920 √ √ √ 0.8219 0.9792 0.7488 0.9929
Tools
Get Citation
Copy Citation Text
Yuanyuan Peng, Haoyang Li, Wen Li, Yuejin Zhang. LTDA‐Mamba: Retinal Vessel Segmentation Based on a Hybrid CNN‐Mamba Network[J]. Acta Optica Sinica, 2025, 45(7): 0717001
Paper Information
Category: Medical optics and biotechnology
Received: Dec. 13, 2024
Accepted: Jan. 16, 2025
Published Online: Mar. 19, 2025
The Author Email: Peng Yuanyuan (pengmi467347713@126.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20