Skip to main content
. 2021 Mar 25;11:6876. doi: 10.1038/s41598-021-86022-7

Table 4.

Contribution of optional contrasts to classifier performance.

Data set Detection type Δ AUC Δ Sensitivity (%) Δ Specificity (%)
Internal Abnormality 0.01 2 1
Acute hemorrhage 0.01  − 2 6
Acute infarct 0.00 0 0
Mass effect 0.01 3 1
External Abnormality 0.00 0 0
Acute hemorrhage 0.01  − 2 3
Acute infarct 0.00 0 0
Mass effect 0.01 2  − 2

Positive values denote that the model performance was higher with all available contrasts than with only mandatory contrasts.