Table 3 Diagnostic performance of non-gated model.
From: Automated coronary calcium scoring using deep learning with multicenter external validation
Cutoff CAC score | Metrica | Internal validation cohorts | External validation cohorts | ||||
|---|---|---|---|---|---|---|---|
Stanford | MESA | Site 1 | Site 2 | Site 3 | Site 4 | ||
Cohen’s Kappa | 0.836 | 0.517 | 0.802 | 0.684 | 0.644 | 0.583 | |
1 | Sensitivity (%) | 100 (100, 100) | 84.6 (74.2, 95.0) | 94.4 (84.9, 100) | 92.5 (86.5, 98.4) | 93.0 (87.0, 99.0) | 81.9 (75.4, 88.4) |
Specificity (%) | 94.4 (87.5, 100) | 100 (100, 100) | 100 (100, 100) | 90.9 (84.4, 97.4) | 78.6 (68.9, 88.3) | 90.4 (85.4, 95.4) | |
PPV (%) | 96.0 (90.1, 100) | 100 (100, 100) | 100 (100, 100) | 96.1 (91.7, 100) | 87.0 (79.0, 94.9) | 93.2 (88.9, 97.4) | |
NPV (%) | 100 (100, 100) | 53.8 (39.4, 68.3) | 80.0 (63.3, 96.7) | 83.3 (74.9, 91.8) | 88.0 (80.3, 95.7) | 75.8 (68.6, 83.0) | |
F1 | 0.980 | 0.917 | 0.971 | 0.942 | 0.899 | 0.872 | |
10 | Sensitivity (%) | 100 (100, 100) | 88.6 (79.4, 97.8) | 100 (100, 100) | 95.9 (91.4, 100) | 92.3 (86.0, 98.6) | 88.9 (83.6, 94.2) |
Specificity (%) | 95.2 (88.8, 100) | 90.9 (82.6, 99.2) | 100 (100, 100) | 88.5 (81.2, 95.7) | 84.4 (75.8, 92.9) | 90.5 (85.5, 95.4) | |
PPV (%) | 95.5 (89.2, 100) | 96.9 (91.8, 100) | 100 (100, 100) | 94.0 (88.6, 99.4) | 87.8 (80.1, 95.5) | 91.4 (86.7, 96.2) | |
NPV (%) | 100 (100, 100) | 71.4 (58.4, 84.5) | 100 (100, 100) | 92.0 (85.9, 98.1) | 90.0 (82.9, 97.1) | 87.7 (82.2, 93.2) | |
F1 | 0.977 | 0.925 | 1.000 | 0.949 | 0.900 | 0.901 | |
100 | Sensitivity (%) | 93.8 (86.4, 100) | 71.4 (58.4, 84.5) | 92.9 (82.1, 100) | 90.6 (84.0, 97.2) | 88.5 (80.9, 96.0) | 74.0 (66.6, 81.4) |
Specificity (%) | 96.2 (90.3, 100) | 92.0 (84.2, 99.8) | 100 (100, 100) | 93.0 (87.3, 98.8) | 93.3 (97.4, 99.2) | 94.1 (90.1, 98.1) | |
PPV (%) | 93.8 (86.4, 100) | 88.2 (78.9, 97.5) | 100 (100, 100) | 90.6 (84.0, 97.2) | 88.5 (80.9, 96.0) | 88.1 (82.6, 93.6) | |
NPV (%) | 96.2 (90.3, 100) | 79.3 (67.6, 91.0) | 88.9 (75.8, 100) | 93.0 (87.3, 98.8) | 93.3 (97.4, 99.2) | 86.0 (80.2, 91.9) | |
F1 | 0.938 | 0.789 | 0.963 | 0.906 | 0.885 | 0.804 | |
400 | Sensitivity (%) | 83.3 (72.1, 94.6) | 75.0 (62.5, 87.5) | 100 (100, 100) | 93.8 (88.3, 99.2) | 91.7 (85.1, 98.2) | 83.3 (77.0, 89.6) |
Specificity (%) | 100 (100, 100) | 100 (100, 100) | 91.7 (80.2, 100) | 93.2 (87.5, 98.9) | 98.3 (95.3, 100) | 97.4 (94.8, 100) | |
PPV (%) | 100 (100, 100) | 100 (100, 100) | 90.9 (78.9, 100) | 78.9 (69.7, 88.2) | 91.7 (85.1, 98.2) | 83.3 (77.0, 89.6) | |
NPV (%) | 97.3 (92.4, 100) | 95.0 (88.7, 100) | 100 (100, 100) | 98.2 (95.2, 100) | 98.3 (95.3, 100) | 97.4 (94.8, 100) | |
F1 | 0.909 | 0.857 | 0.952 | 0.857 | 0.917 | 0.802 | |
