. 2022 Mar 15;12:4433. doi: 10.1038/s41598-022-07843-8

Table 1.

The six evaluation metrics are chosen from the conventional and well know evaluation parameters presented in the MICCAI grand challenge and literature. The Dice index (DI), Jaccard index (JI), and volume difference (VD) are used to measure the quality of WML segmentation by calculating their volume and Euclidean distance of WML boundary. Whereas, the true positive rate (TPR), false positive rate (FPR) and positive predictive value (PPV) usually practice understanding the quality of WML detection.

Evaluation metrics	Definition	Best	Worst
DI	$\frac{2 \times V o l (S_{auto}) \cap V o l (G T)}{V o l (S_{auto}) + V o l (G T)}$	1	0
JI	$\frac{2 \times V o l (S_{auto}) \cap V o l (G T)}{V o l (S_{auto}) \cup V o l (G T)}$	1	0
TPR	$\frac{TP}{T P + F N}$	1	0
FPR	$\frac{FP}{T P + T N}$	0	1
PPV	$\frac{TP}{T P + F P}$	1	0
VD	$\frac{V o l (S_{auto}) - V o l (G T)}{V o l (G T)}$	0	∞

The notations for Vol ( $S_{auto}$ ) is volume of WML segmented by automated method, Vol ( $GT$ ) is volume of WML delineated by neuroradiologist, TP is number of true positive, FP is number of false positive, FN is number of false negative, and TN is number of true negative. The best is defined as perfect segmentation and worst is completely missed segmentation.