Table 4.

Evaluation results by using different numbers of weakly labeled subjects with and without postprocessing. In particular, the mean dice coefficient is presented in the way of mean ± standard deviation.

Scale of the dataset	Postprocessing	DC	P L	R L	F1	DR
398 subjects	No	0.651 ± 0.158	0.403	0.956	0.567	1.000
318 subjects		0.649 ± 0.157	0.391	0.949	0.554	1.000
238 subjects		0.630 ± 0.165	0.383	0.949	0.546	1.000
159 subjects		0.593 ± 0.184	0.297	0.949	0.452	1.000
79 subjects		0.620 ± 0.209	0.487	0.898	0.632	0.979
398 subjects	Yes	0.699 ± 0.128	0.852	0.923	0.886	1.000
318 subjects		0.700 ± 0.134	0.821	0.920	0.867	1.000
238 subjects		0.675 ± 0.138	0.843	0.901	0.871	1.000
159 subjects		0.665 ± 0.181	0.854	0.872	0.863	0.966
79 subjects		0.557 ± 0.250	0.793	0.741	0.766	0.922