Table 3 Comparison of performance metrics for various models in Lead II for cardiovascular disease classification. (Highest performance values are in bold.)

From: Integrating snapshot ensemble learning into masked autoencoders for efficient self-supervised pretraining in medical imaging

 

AUC

AUPRC

Accuracy (%)

Sensitivity

Precision

F1-score

ResNet-34 (FS)

0.990

0.936

89.66

0.897

0.890

0.891

ViT-S (FS)

0.975

0.897

84.10

0.841

0.833

0.823

ResNet-34 (IN)

0.991

0.938

90.23

0.902

0.889

0.892

ViT-S (IN)

0.982

0.918

88.60

0.886

0.878

0.879

MAE (Scalo)

0.992

0.954

91.86

0.919

0.914

0.914

Snap-MAE (Scalo)

0.994

0.964

93.49

0.935

0.934

0.930

Back to article page