Table 2 Model performance for predicting neonatal respiratory morbidity depending on different features.
Clinical features | Radiomics features | Clinical and radiomics features | ||||
|---|---|---|---|---|---|---|
Training set (95% CI) | Testing set (95% CI) | Training set (95% CI) | Testing set (95% CI) | Training set (95% CI) | Testing set (95% CI) | |
Sensitivity | 45.10% (38.27–51.99%) | 55.56% (43.35–67.96%) | 74.51% (68.58–80.27%) | 66.67% (54.91–78.86%) | 84.31% (79.06–89.44%) | 77.78% (68.30–87.43%) |
Specificity | 90.57% (88.23–92.89%) | 83.58% (79.10–88.00%) | 73.58% (70.11–77.02%) | 74.63% (69.30–80.08%) | 81.13% (78.16–84.07%) | 82.09% (77.65–86.62%) |
Accuracy | 79.52% (76.60–82.43%) | 77.65% (73.20–82.10%) | 73.81% (70.80–76.74%) | 72.94% (68.06–77.98%) | 81.90% (79.34–84.41%) | 81.18% (77.33–85.12%) |
PPV | 60.53% (52.29–68.82%) | 47.62% (36.07–59.01%) | 47.50% (42.00–52.93%) | 41.38% (32.04–50.99%) | 58.90% (53.25–64.51%) | 53.85% (44.08–63.93%) |
NPV | 83.72% (80.81–86.61%) | 87.50% (83.43–91.67%) | 90.00% (87.46–92.47%) | 89.29% (85.03–93.66%) | 94.16% (92.08–96.17%) | 93.22% (90.16–96.31%) |
AUC | 0.70 (0.65–0.75) | 0.61 (0.52–0.70) | 0.79 (0.75–0.84) | 0.67 (0.58–0.76) | 0.88 (0.83–0.92) | 0.83 (0.79–0.97) |
