Table 10 Effect of different attention mechanisms for mean AUC Results on the ChestX-Ray14 dataset. Significant values are in bold.
From: Automated thorax disease diagnosis using multi-branch residual attention network
Attention | Atel | Card | Effu | Infi | Mass | Nodu | pneu1 | pneu2 | Cons | Edem | Emph | Fibr | P_T | Hern | Mean |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
SE | 0.819 | 0.869 | 0.866 | 0.692 | 0.850 | 0.785 | 0.761 | 0.840 | 0.795 | 0.881 | 0.916 | 0.854 | 0.774 | 0.925 | 0.831 |
CBAM | 0.820 | 0.870 | 0.867 | 0.692 | 0.845 | 0.778 | 0.767 | 0.838 | 0.789 | 0.884 | 0.903 | 0.846 | 0.776 | 0.967 | 0.832 |
CA | 0.830 | 0.877 | 0.881 | 0.705 | 0.860 | 0.792 | 0.766 | 0.866 | 0.816 | 0.891 | 0.927 | 0.847 | 0.789 | 0.922 | 0.841 |
