Table 3.

Nucleus segmentation model DICE results comparing to other methods. * indicates that the results were obtained from Ref. 5.

	Methods	TNBC	All CoNSeP	CoNSeP test
Non-deep learning	Cell Profiler*	—	—	0.434
	QuPath*	—	—	0.588
Deep Learning	FCN8 + WS*	0.726	0.609	—
	SegNet + WS*	0.758	0.681	—
	U-Net*	0.681	0.585	—
	Mask-RCNN*	0.705	0.606	—
	DCAN*	0.725	0.609	—
	Micro-Net*	0.701	0.644	—
	DIST*	0.719	0.621	—
	HoVer-Net*	0.749	0.664	—
	NuKit	0.793	0.854	0.860