Skip to main content
. 2022 Dec 1;37:103278. doi: 10.1016/j.nicl.2022.103278

Table 4.

Comparison of the deep learning model and ADC-thresholding model performance in all patients using a probability threshold of 0.5.

Metrics’ median (IQR) AUC Sensitivity Specificity DSC Volume error, (ml) Absolute volume error, (ml)
DCNN 0.91
(0.84–0.96)
0.60
(0.16–0.84)
0.97
(0.93–0.99)
0.50
(0.17–0.66)
0
(–22–30)
27
(7–60)
ADC-thresholding 0.62
(0.56–0.69)
0.26
(0.12–0.40)
0.98
(0.97–0.99)
0.18
(0.10–0.35)
7
(–54–66)
64
(25–96)
P-value < 0.001 < 0.001 < 0.001 < 0.001 0.50 < 0.001