• Opto-Electronic Engineering
  • Vol. 50, Issue 1, 220199 (2023)
Liming Liang*, Xin Dong, Renjie Li, and Anjun He
Author Affiliations
  • School of Electrical Engineering and Automation, Jiangxi University of Science and Technology, Ganzhou, Jiangxi 341000, China
  • show less
    DOI: 10.12086/oee.2023.220199 Cite this Article
    Liming Liang, Xin Dong, Renjie Li, Anjun He. Classification algorithm of retinopathy based on attention mechanism and multi feature fusion[J]. Opto-Electronic Engineering, 2023, 50(1): 220199 Copy Citation Text show less
    Structure of the ResNeSt block
    Fig. 1. Structure of the ResNeSt block
    Multi-scale feature boosting module
    Fig. 2. Multi-scale feature boosting module
    Graphic feature fusion module
    Fig. 3. Graphic feature fusion module
    Overall framework of the algorithm
    Fig. 4. Overall framework of the algorithm
    Comparison of fundus images before and after preprocessing in different DR stages. (a) Original image;(b) Preprocessed image
    Fig. 5. Comparison of fundus images before and after preprocessing in different DR stages. (a) Original image;(b) Preprocessed image
    AMMF network feature heatmap
    Fig. 6. AMMF network feature heatmap
    Training loss curves on IDRID and Eye-PACS datasets
    Fig. 7. Training loss curves on IDRID and Eye-PACS datasets
    Confusion matrix. (a) Ref. [20]; (b) Ours
    Fig. 8. Confusion matrix. (a) Ref. [20]; (b) Ours
    Comparison trend graph. (a) AUC value of each category of DR; (b) Comparison of each indicator bar chart
    Fig. 9. Comparison trend graph. (a) AUC value of each category of DR; (b) Comparison of each indicator bar chart
    ModelQWK/%Acc/%Se/%Sp/%
    M184.8371.8488.4888.23
    M285.0874.7589.8094.11
    M387.7077.6692.7589.80
    M489.5976.6996.1088.23
    AMMF90.3878.6495.6591.17
    Table 1. Performance of ablation experiments in the IDRID dataset
    MethodsBackboneQWK/%Acc/%Se/%Sp/%
    Bhardwaj et al.[18]ResNet-5079.4682.8576.98
    Wu et al.[19]ResNet-1856.1964.2187.39
    Song et al.[20]ResNet-5086.5476.6989.8588.00
    Shi et al.[21]Efficientnet-b587.6379.06
    OursResNeSt-5090.3878.6495.6591.17
    Table 2. Performance of different models in the IDRID dataset
    MethodsBackboneAcc/%AUC/%Se/%Sp/%
    Wu et al.[19]ResNet-1883.1089.0053.9991.22
    Song et al.[20]ResNet-5082.2587.0861.2790.21
    Kamothi et al.[22]Modified VGG-1677.6077.00
    Thota et al.[23]VGGNet74.0080.0080.0065.00
    Saxena et al.[24]InceptionResNetV276.6879.5150.7791.25
    OursResNeSt-5084.4190.3677.8597.05
    Table 3. Performance of different models in the Eye-PACS dataset
    Liming Liang, Xin Dong, Renjie Li, Anjun He. Classification algorithm of retinopathy based on attention mechanism and multi feature fusion[J]. Opto-Electronic Engineering, 2023, 50(1): 220199
    Download Citation