Table 1.

The quantitative evaluation of the proposed method and the baseline approaches in the segmentation of metastases on H–E WSIs.

	Method	Score			95% C.I. for Mean
		Mean	Std. Deviation	Std. Error	Lower Bound	Upper Bound
	Proposed method	0.892	0.163	0.047	0.787	0.995
Precision	U-Net [16]	0.486	0.116	0.033	0.411	0.559
	SegNet [15]	0.548	0.091	0.026	0.489	0.605
	FCN [17]	0.552	0.062	0.018	0.512	0.590
	Deeplabv3+ [18] with MobileNet [19]	0.643	0.262	0.075	0.476	0.809
	Deeplabv3+ [18] with ResNet [20]	0.613	0.354	0.102	0.388	0.838
	Deeplabv3+ [18] with Xception [21]	0.753	0.286	0.082	0.571	0.935
	Proposed method	0.837	0.169	0.049	0.729	0.945
Recall	U-Net [16]	0.643	0.022	0.006	0.628	0.656
	SegNet [15]	0.588	0.028	0.008	0.570	0.606
	FCN [17]	0.500	0.082	0.023	0.448	0.551
	Deeplabv3+ [18] with MobileNet [19]	0.682	0.277	0.080	0.506	0.858
	Deeplabv3+ [18] with ResNet [20]	0.440	0.261	0.075	0.274	0.606
	Deeplabv3+ [18] with Xception [21]	0.584	0.290	0.083	0.399	0.768
	Proposed method	0.844	0.127	0.036	0.763	0.925
F1-score	U-Net [16]	0.564	0.095	0.027	0.503	0.624
	SegNet [15]	0.562	0.124	0.036	0.383	0.581
	FCN [17]	0.510	0.078	0.022	0.400	0.531
	Deeplabv3+ [18] with MobileNet [19]	0.640	0.241	0.069	0.487	0.794
	Deeplabv3+ [18] with ResNet [20]	0.480	0.262	0.075	0.313	0.646
	Deeplabv3+ [18] with Xception [21]	0.621	0.259	0.047	0.456	0.786
	Proposed method	0.749	0.188	0.054	0.629	0.868
mIoU	U-Net [16]	0.473	0.114	0.331	0.400	0.546
	SegNet [15]	0.380	0.129	0.037	0.298	0.462
	FCN [17]	0.363	0.086	0.025	0.308	0.418
	Deeplabv3+ [18] with MobileNet [19]	0.504	0.229	0.066	0.358	0.650
	Deeplabv3+ [18] with ResNet [20]	0.344	0.213	0.031	0.208	0.480
	Deeplabv3+ [18] with Xception [21]	0.487	0.251	0.072	0.327	0.647
	v3_DCNN-1280 * [22]	0.685	-	-	-	-
	Xception-65 * [23]	0.645	-	-	-	-

* The reported numbers of Guo [22] and Priego [23] are referred in this table.