Table 2 Performances of the rule-based models developed to automatically annotate the texts of the radiological reports
Variable investigated with text mining | Accuracy | Error rate | Sensitivity | Specificity | PPV | NPV | AUROC |
|---|---|---|---|---|---|---|---|
GGO positive | 98 (93–99.8) | 2% | 96 (86.3–99.5) | 100 (92.9–100) | 93.9 (83.6–97.2) | 97.3 (86.6–99.1) | 0.98 (0.95–1) |
Reticulation positive | 99 (94.6–100) | 1% | 98 (89.4–99.9) | 100 (92.9–100) | 95.6 (85.3–98.2) | 97.4 (86.8–99.1) | 0.99 (0.97–1) |
Bilateral positive | 98 (93–99.8) | 2% | 96.2 (86.8–99.5) | 100 (92.6–100) | 93.6 (83–97.1) | 97.4 (87.1–99.1) | 0.98 (0.95–1) |
Subpleural positive | 93 (86.1–97.1) | 7% | 96 (86.3–99.5) | 90 (78.2–96.7) | 95.7 (85.2–98.9) | 90.6 (80.7–95.7) | 0.93 (0.89–0.98) |