Table 2.

Studies using deep learning approach for bladder cancer diagnosis and staging.

Author	Year	Modality	Number of patients (Train/Val/Test)	CNN structure	Performance (validation or testing dataset)
Yang et al. (52)	2021	CT	369 patients,1200 images (70%/15%/15%)	A small convolutional network contains four conv_layer+max_pooling_layer stages/eight pretrained models, 2D	Accuracy (small) = 0.861 AUROC (small) = 0.998 Accuracy (VGG16) = 0.939 AUROC (VGG16) = 0.997
Zhang et al. (53)	2021	CT	183/110/73 (internal)/75 (external)	FGP-Net (a novel convolutional network contains Dense Blocks and DFL modules), 3D	AUC (internal) = 0.861 Accuracy (internal) = 0.795 AUC (external) = 0.791 Accuracy (external) = 0.747
Liu et al. (54)	2022	T2W MRI	51/8/16	ResNet18 with the super-resolution module and the Non-local attention module, 2D	Sensitivity = 94.74
Taguchi et al. (55)	2021	T2W MRI	68	The denoising Deep Learning Reconstruction (dDLR)	–

AUC, area under curve; Sensitivity=TP/(TP+ FN).