Table 5 Diagnostic performance of the clinical model, radiomics nomogram, and readers in the training and validation cohorts
AUC | Accuracy | Sensitivity | Specitivity | |
|---|---|---|---|---|
Training cohort | ||||
Radiomics nomogram | 0.923 (0.876, 0.971) | 84.2 (83.9, 84.4) | 97.4 (92.5, 100) | 75.8 (65.1, 86.5) |
Clinical model | 0.723 (0.630, 0.817) | 68.3 (67.9, 68.7) | 61.5 (46.3, 76.8) | 72.6 (61.5, 83.7) |
Pooled readers | 0.621 (0.523, 0.719) | 63.4 (62.9, 63.8) | 56.4 (40.8, 72.0) | 67.7 (56.1, 79.4) |
Radiologist 1 | 0.617 (0.519, 0.716) | 62.4 (61.9, 62.8) | 59.0 (43.5, 74.4) | 64.5 (52.6, 76.4) |
Radiologist 2 | 0.664 (0.574, 0.753) | 63.4 (62.9, 63.8) | 79.5 (66.8, 92.2) | 53.2 (40.8, 65.6) |
Validation cohort | ||||
Radiomics nomogram | 0.888 (0.792, 0.983) | 84.0 (83.5, 84.5) | 73.1 (56.0, 90.1) | 95.8 (87.8, 100) |
Clinical model | 0.754 (0.617, 0.891) | 72.0 (71.2, 72.8) | 65.4 (47.1, 83.7) | 79.2 (62.9, 95.4) |
Pooled readers | 0.647 (0.524, 0.771) | 64.0 (63.1, 64.9) | 46.2 (27.0, 65.3) | 83.3 (68.4, 98.2) |
Radiologist 1 | 0.623 (0.489, 0.758) | 62.0 (61.1, 62.9) | 53.8 (34.7, 73.0) | 70.8 (52.6, 89.0) |
Radiologist 2 | 0.659 (0.525, 0.793) | 66.0 (65.1, 66.9) | 69.2 (51.5, 87.0) | 62.5 (43.1, 81.9) |