TABLE 2.

Comparison of F1 scores of different methods applied to the TCGA test set.

Organ	Bladder	Colorectal	Stomach	Breast	Kidney	Liver	Prostate	Overall
FCN-8 (Long et al., 2015)	0.8084	0.6934	0.7982	0.8113	0.5797	0.7589	0.8367	0.7552
Mask R-CNN (He et al., 2015)	0.7610	0.6820	0.8268	0.7481	0.7554	0.7157	0.7401	0.7470
U-Net (Ronneberger et al., 2015)	0.7953	0.7360	0.8638	0.7818	0.7913	0.6981	0.7904	0.7795
CNN3 (Kumar et al., 2017)	0.7808	0.7399	0.8948	0.7181	0.7222	0.6881	0.7922	0.7623
DIST (Naylor et al., 2019)	0.8196	0.7286	0.8534	0.8071	0.7706	0.7281	0.7967	0.7863
Stacked U-Net	0.8249	0.7685	0.8498	0.7990	0.7986	0.7276	0.7829	0.7930
U-Net (DLA)	0.8296	0.7756	0.8530	0.8025	0.7994	0.7296	0.7895	0.7970
A two-stage U-Net (Mahbod et al., 2019)	0.7599	0.7668	0.8912	0.8024	0.8531	0.7938	0.8648	0.8189
Two-stage learning U-Net (DLA) (Kang et al., 2019)	0.8360	0.7808	0.8629	0.8183	0.8022	0.7513	0.8037	0.8079
Ours	0.8217	0.8135	0.8690	0.8123	0.8251	0.7865	0.8451	0.8247

The best results are shown in bold.