Skip to main content

Table 3 Performances of WSDL model and HAT and SEDAN score

From: A weakly supervised deep learning model integrating noncontrasted computed tomography images and clinical factors facilitates haemorrhagic transformation prediction after intravenous thrombolysis in acute ischaemic stroke patients

 

AUC

Accuracy

Sensitivity

Specificity

PPV

NPV

WSDL

0.799 (0.712,0.883)

0.735 (0.683,0.787)

0.797 (0.632,0.955)

0.730 (0.674,0.784)

0.213 (0.163,0.263)

0.976 (0.956,0.994)

HAT

0.753 (0.697,0.807)

0.797 (0.771,0.824)

0.551 (0.438,0.66)

0.819 (0.793,0.847)

0.217 (0.174,0.260)

0.953 (0.941,0.964)

SEDAN

0.777 (0.721,0.837)

0.673 (0.642,0.704)

0.768 (0.670,0.872)

0.664 (0.632,0.697)

0.172 (0.150,0.196)

0.969 (0.957,0.983)