Table 3 Comparison results for four-class classification
From: Classification and visual explanation for COVID-19 pneumonia from CT images using triple learning
Tasks | Accuracy (%) | Precision (%) | Sensitivity (%) | Specificity (%) | F-measure (%) |
|---|---|---|---|---|---|
3D network | |||||
3D-ResNet1852 | 45.39±3.16 | 36.91±2.58 | 36.82±3.65 | 44.84±2.97 | 35.06±2.73 |
CovNet (ResNet18)50 | 56.81±0.23 | 45.78±2.20 | 43.26±5.53 | 51.61±6.25 | 41.51±4.25 |
DeCovNet51 | 50.63±3.79 | 45.42±2.69 | 45.30±2.11 | 49.99±3.65 | 44.58±2.56 |
2D network | |||||
Baseline (ResNet18) | 49.48±2.50 | 42.05±1.34 | 41.11±1.71 | 48.67±2.54 | 40.82±1.48 |
WSDL4 | 53.66±1.48 | 44.04±0.80 | 45.61±2.80 | 53.13±1.37 | 42.63±0.84 |
ABN44 | 51.10±0.67 | 42.26±0.96 | 41.81±1.62 | 50.34±0.65 | 41.32±1.07 |
MTDL7 | 52.38±3.36 | 42.31±2.39 | 40.49±3.83 | 47.38±6.95 | 40.42±2.96 |
Double Net (ours) | 51.11±1.52 | 41.23±1.63 | 39.62±1.75 | 50.35±1.61 | 39.87±1.65 |
Triple Net (ours) | 54.02±2.30 | 43.84±2.63 | 40.37±4.56 | 44.51±6.40 | 41.70±3.71 |
Triple Net + WSDL4(ours) | 58.22±3.35 | 47.22±3.06 | 48.82±4.69 | 57.89±3.31 | 45.30±3.65 |
