Table 2 Performance of AI classification of texture and morphological features utilizing five different classifiers of COVID-19 (N = 130) vs bacterial pneumonia (N = 145)
Classifier | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) | Accuracy (%) | AUC (LB, UP) | p value |
|---|---|---|---|---|---|---|---|
XGB-L | 95.35 | 97.44 | 97.62 | 95.00 | 96.34 | 0.98 (0.96, 1.00) | 2.19e−14 |
XGB-Tree | 88.37 | 94.87 | 95.00 | 88.10 | 91.46 | 0.97 (0.93, 1.00) | 2.19e−14 |
CART (DT) | 74.42 | 97.44 | 96.97 | 77.55 | 85.37 | 0.92 (0.87, 0.98) | 3.22e−10 |
KNN | 86.05 | 56.41 | 68.52 | 78.57 | 71.95 | 0.83 (0.74, 0.92) | 2.40e−04 |
Naïve Bayes | 90.70 | 79.49 | 82.98 | 88.51 | 85.37 | 0.92 (0.85, 0.98) | 3.22e−10 |