Table 5.

Ablation study of the contribution of each component in our proposed network. Data in parentheses are 95% CIs.

Methods	AUC	Accuracy	$F1$-score
$L_{\text{Swin}-\mathrm{T}}$	0.812 (0.799, 0.826)	0.749 (0.740, 0.757)	0.692 (0.670, 0.714)
$L_{\text{fusion}}$	0.851 (0.844, 0.858)	0.757 (0.746, 0.767)	0.726 (0.707, 0.745)
$L_{\text{fusion}}+L_{\text{Swin}-\mathrm{T}}+L_{\mathrm{CNN}}+L1$	0.856 (0.846, 0.866)	0.755 (0.747, 0.762)	0.717 (0.703, 0.729)