Table 2 Results of the classification and regression networks on the evaluation set (\(n=7143\)).

From: Emphysema subtyping on thoracic computed tomography scans using deep neural networks

Method

Subtype

ACC(%)

F-measure

Linear weighted kappa (95% CI)

The Humphries algorithm10

Centrilobular

45

–

60

Ours (classification)

Centrilobular

52.23

51.00

64.29 (63.16–65.42)

Ours (classification)

Paraseptal

59.12

57.12

42.03 (40.21–43.85)

Ours (regression)

Centrilobular

51.32

49.61

64.24 (63.14–65.35)

Ours (regression)

Paraseptal

64.62

60.74

52.06 (50.40–53.73)

  1. The classification accuracy (ACC), F-measurements (F-measure), and linear weighted kappa are calculated against the visual scores. We also list the results from the Humphries algorithm, obtained on the same test set, where only ACC and kappa were reported for predicting centrilobular emphysema severity scores.
Back to article page