Optics and Precision Engineering, Volume. 32, Issue 5, 714(2024)
DRT Net： dual Res-Transformer pneumonia recognition model oriented to feature enhancement
Figures & Tables(14)
Fig. 10. Radar chart of pneumonia classification results for different models
Fig. 11. Confusion matrix of classification results for each model
Table 1. Result comparison of ablation experiments
View tableView in ArticleTable 1. Result comparison of ablation experiments
Name Model GADRM GLFEM CDAFFM A P R F1 AUC Network1 DPN92 0.969 2 0.912 8 0.902 6 0.900 3 0.988 4 Network2 DPN92+CDAFFM √ 0.972 5 0.927 0 0.915 5 0.919 0 0.990 1 Network3 DPN92+GLFEM √ 0.974 1 0.928 8 0.920 5 0.921 1 0.991 2 Network4 DPN92+GADRM √ 0.974 7 0.936 3 0.942 7 0.938 6 0.992 8 Network5 DPN92+GLFEM+CDAFFM √ √ 0.974 6 0.930 9 0.913 8 0.920 0 0.992 7 Network6 DPN92+GADRM+GLFEM √ √ 0.975 5 0.923 5 0.917 3 0.915 3 0.992 0 Network7 DPN92+GADRM+CDAFFM √ √ 0.977 9 0.939 1 0.937 6 0.937 4 0.993 0 Network8 DRT Net √ √ √ 0.984 1 0.944 2 0.942 0 0.942 6 0.996 5
Table 2. Classification results of pneumonia X-ray images for each model
View tableView in ArticleTable 2. Classification results of pneumonia X-ray images for each model
Model A P R F1 AUC ResNet50［ 5 ］0.959 6 0.914 2 0.907 4 0.905 1 0.986 9 ResNet101［ 5 ］0.965 1 0.927 0 0.919 4 0.920 3 0.991 0 Res2Net50［ 18 ］0.966 4 0.907 5 0.918 4 0.908 8 0.990 4 Densenet121［ 19 ］0.969 1 0.929 0 0.912 5 0.916 7 0.991 2 ResNeXt101［ 20 ］0.963 3 0.921 2 0.918 7 0.917 4 0.990 8 MobileNetV2［ 21 ］0.964 7 0.913 9 0.894 4 0.897 1 0.989 4 DPN92［ 22 ］0.969 2 0.912 8 0.902 6 0.900 3 0.988 4 Swin Transformer［ 23 ］0.940 3 0.900 7 0.868 4 0.876 2 0.979 1
Table 2. Classification results of pneumonia X-ray images for each model
View tableView in ArticleTable 2. Classification results of pneumonia X-ray images for each model
Model A P R F1 AUC Mobile vit［ 24 ］0.974 1 0.930 7 0.918 4 0.922 3 0.988 6 ReNeSt50［ 25 ］0.961 8 0.910 5 0.902 8 0.904 1 0.982 7 ConvNeXt_XL［ 26 ］0.966 4 0.893 0 0.854 6 0.854 3 0.989 4 DRT Net 0.984 1 0.944 2 0.942 0 0.942 6 0.996 5
Tools
Get Citation
Copy Citation Text
Tao ZHOU, Caiyue PENG, Yuhu DU, Pei DANG, Fengzhen LIU, Huiling LU. DRT Net： dual Res-Transformer pneumonia recognition model oriented to feature enhancement[J]. Optics and Precision Engineering, 2024, 32(5): 714
Paper Information
Category:
Received: May. 11, 2023
Accepted: --
Published Online: Apr. 2, 2024
The Author Email: Caiyue PENG (peng_caiyue@163.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20