Skip to main content
European Radiology Experimental

Table 4 Performance metrics on the local test data using early stopping from RSPECT (experiment A) or from a small sample of 15 CTPA exams (local data)

From: Leveraging open dataset and transfer learning for accurate recognition of chronic pulmonary embolism from CT angiogram maximum intensity projection images

Early stopping set

Experiment A early stopping

Local data early stopping

Single model (mean ± SD)

Single model (mean ± SD)

Ensemble model

AUC, all

0.82 ± 0.01

0.86 ± 0.04

0.89

BAcc (ad hoc), alla

0.69 ± 0.04

0.74 ± 0.07

0.75

BAcc (post hoc), allb

0.76 ± 0.02

0.81 ± 0.03

0.82

AUC, LRmax

0.87 ± 0.01

0.89 ± 0.04

0.94

BAcc (ad hoc), LRmaxa

0.67 ± 0.06

0.76 ± 0.07

0.87

BAcc (post hoc), LRmaxb

0.82 ± 0.02

0.83 ± 0.05

0.89

  1. The operating points for the test set balanced accuracy calculations were chosen by selecting the threshold nearest to the top-left corner of the receiver operator characteristic curve calculated from the early stopping set (a) or from the final test set (b). AUC Area under the receiver operator characteristic curve, BAcc Balanced accuracy, LRmax Maximum of the left and right lung prediction, SD Standard deviation
Back to article page