Table 4.
Accuracy and predictive value between those models.
| Training dataset | AUC | 95%CI | Sensitivity | Specificity | Accuracy | PPV | NPV |
|---|---|---|---|---|---|---|---|
| CT morphological features | 0.877 | 0.819–0.935 | 83.6% (61/73) | 79.4% (50/63) | 81.6% (111/136) | 82.4% (61/74) | 80.7% (50/62) |
| Single-slice histogram | 0.872 | 0.813–0.931 | 86.3% (63/73) | 76.2% (48/63) | 81.6% (111/136) | 80.8% (63/78) | 82.8% (48/58) |
| Whole-lesion histogram | 0.888 | 0.830–0.946 | 82.2% (60/73) | 87.3% (55/63) | 84.6% (115/136) | 88.2% (60/68) | 80.9% (55/68) |
| Morphological-histogram | 0.919 | 0.871–0.968 | 84.9% (62/73) | 92.1% (58/63) | 88.2% (120/136) | 92.5% (62/67) | 84.1% (58/69) |
| Validation dataset | AUC | 95%CI | Sensitivity | Specificity | Accuracy | PPV | NPV |
| CT morphological features | 0.823 | 0.708–0.938 | 78.5% (22/28) | 75% (24/32) | 76.7% (46/60) | 73.3% (22/30) | 80% (24/30) |
| Single-slice histogram | 0.819 | 0.702–0.936 | 75% (21/28) | 71.9% (23/32) | 73.3% (44/60) | 70% (21/30) | 76.7% (23/30) |
| Whole-lesion histogram | 0.865 | 0.773–0.957 | 75% (21/28) | 84.4% (27/32) | 80% (48/60) | 80.8% (21/26) | 79.4% (27/34) |
| Morphological-histogram | 0.895 | 0.813–0.977 | 78.5% (22/28) | 84.4% (27/32) | 81.7% (49/60) | 81.5%(22/27) | 81.8%(27/33) |
CI, confidence interval; AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value.
p < 0.05.