TABLE 3.

Quantitative comparison of different methods applied to the TNBC dataset.

Organ	Recall	Precision	F1-Score	AJI
DeconvNet (Noh et al., 2015)	0.773	0.864	0.805	–
FCN-8 (Long et al., 2015)	0.752	0.823	0.763	–
U-Net (Ronneberger et al., 2015)	0.800	0.820	0.810	0.578
Ensemble (Naylor et al., 2017)	0.900	0.741	0.802	–
Stacked U-Net	0.802	0.830	0.816	0.580
U-Net (DLA)	0.812	0.826	0.818	0.586
DIST (Naylor et al., 2019)	–	–	0.824	0.585
Two-stage learning U-Net (DLA) (Kang et al., 2019)	0.833	0.826	0.829	0.611
Ours	0.853	0.792	0.806	0.621

The best results are shown in bold.