Table 2.
Performance metrics for the small-nodule radiomics predictive vector, diameter and volume models.
| AUC | Accuracy | Sens | Spec | PPV | NPV | |
|---|---|---|---|---|---|---|
| Training | ||||||
| SN-RPV | 0.85 (0.82–0.88) | 78% (74–81%) | 0.86 | 0.69 | 0.74 | 0.83 |
| Diameter | 0.79 (0.75–0.82) | 74% (70–77%) | 0.75 | 0.73 | 0.71 | 0.76 |
| Volume | 0.78 (0.75–0.82) | 73% (69–77%) | 0.80 | 0.66 | 0.70 | 0.77 |
| Test | ||||||
| SN-RPV | 0.78 (0.70–0.86) | 73% (65–81%) | 0.86 | 0.56 | 0.72 | 0.76 |
| Diameter | 0.69 (0.60–0.78) | 64% (56–72%) | 0.65 | 0.64 | 0.73 | 0.55 |
| Volume | 0.77 (0.68–0.85) | 77% (69–84%) | 0.83 | 0.69 | 0.78 | 0.76 |
| External | ||||||
| SN-RPV | 0.78 (0.58–0.92) | 75% (69–80%) | 0.63 | 0.79 | 0.50 | 0.87 |
| Diameter | 0.68 (0.61–0.76) | 70% (64–75%) | 0.38 | 0.80 | 0.39 | 0.80 |
| Volume | 0.80 (0.73–0.86) | 69% (62–74%) | 0.83 | 0.64 | 0.43 | 0.92 |
95% confidence intervals are displayed in brackets. SN-RPV radiomics predictive vector, AUC area under the curve, PPV positive predictive value, NPV negative predictive value.