TABLE 4.

Performances of all algorithms in the subgroups based on WMH load (10 subjects with lowest vs. 10 subjects with highest) compared to the reference segmentation at volume- and lesion- level.

Method	Subgroup	Whole brain volumes			Lesions
		DSC	Sensitivity	Precision	F1-score	Sensitivity	Precision
BIANCA	Low	0.582 ± 0.123	0.668 ± 0.170	0.532 ± 0.126	0.522 ± 0.137	0.590 ± 0.166	0.493 ± 0.157
	High	0.752 ± 0.096	0.668 ± 0.157	0.894 ± 0.060	0.593 ± 0.103	0.575 ± 0.125	0.642 ± 0.132
LST LGA	Low	0.444 ± 0.192	0.416 ± 0.249	0.563 ± 0.098	0.482 ± 0.120	0.379 ± 0.120	0.691 ± 0.134
	High	0.690 ± 0.089	0.637 ± 0.146	0.785 ± 0.097	0.415 ± 0.105	0.348 ± 0.115	0.585 ± 0.195
LST LPA FLAIR only	Low	0.610 ± 0.111	0.631 ± 0.234	0.674 ± 0.148	0.569 ± 0.144	0.528 ± 0.205	0.703 ± 0.206
	High	0.708 ± 0.125	0.763 ± 0.183	0.716 ± 0.183	0.469 ± 0.143	0.454 ± 0.145	0.563 ± 0.216
LST LPA T1w + FLAIR	Low	0.524 ± 0.154	0.518 ± 0.267	0.660 ± 0.112	0.590 ± 0.184	0.500 ± 0.237	0.839 ± 0.137
	High	0.680 ± 0.116	0.721 ± 0.207	0.705 ± 0.157	0.454 ± 0.162	0.408 ± 0.146	0.582 ± 0.218
pgs	Low	0.482 ± 0.202	0.432 ± 0.211	0.615 ± 0.158	0.502 ± 0.144	0.501 ± 0.203	0.557 ± 0.169
	High	0.773 ± 0.133	0.717 ± 0.180	0.883 ± 0.059	0.713 ± 0.159	0.680 ± 0.155	0.751 ± 0.166
sysu_media (default)	Low	0.442 ± 0.202	0.438 ± 0.210	0.457 ± 0.199	0.439 ± 0.150	0.537 ± 0.274	0.395 ± 0.143
	High	0.766 ± 0.107	0.743 ± 0.164	0.826 ± 0.075	0.633 ± 0.119	0.654 ± 0.128	0.623 ± 0.133
sysu_media (retrained)	Low	0.608 ± 0.120	0.623 ± 0.212	0.648 ± 0.085	0.635 ± 0.125	0.573 ± 0.194	0.759 ± 0.103
	High	0.753 ± 0.081	0.712 ± 0.147	0.829 ± 0.090	0.637 ± 0.099	0.591 ± 0.111	0.727 ± 0.165

Mean and standard deviation are displayed. DSC, Sørensen-Dice similarity coefficient; T1w, T1-weighted image; FLAIR, fluid-attenuated inversion recovery image; BIANCA, brain intensity abnormality classification algorithm; LST LGA/-LPA, lesion segmentation toolbox for SPM using lesion growth algorithm/lesion prediction algorithms.