Table 3 The quantitative evaluation of pneumothorax diagnosis results with different models
Models | Accuracy (%) | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | \(F_1\)-score (%) |
|---|---|---|---|---|---|---|
U-Net [18] | 88.39 | 81.16 | 95.31 | 94.31 | 84.07 | 87.24 |
SegNet [32] | 90.78 | 85.87 | 95.49 | 94.80 | 87.58 | 90.11 |
DeepLab v3+ [33] | 91.10 | 85.60 | 96.37 | 95.76 | 87.47 | 90.40 |
DenseASPP [34] | 91.46 | 86.33 | 96.37 | 95.80 | 88.04 | 90.82 |
FC-DenseNet [25] | 92.18 | 84.76 | 99.29 | 99.14 | 87.18 | 91.39 |
MS_scSE_U-Net | 90.96 | 85.96 | 95.75 | 95.10 | 87.68 | 90.30 |
MS_scSE_FC-DenseNet (Ours) | 93.45 | 88.55 | 98.14 | 97.86 | 89.94 | 92.97 |