Table 2. Diagnostic performance of AI on a per-patient and per-vessel basis.
| Dataset | Stenosis | Basis | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) |
|---|---|---|---|---|---|---|
| Internal (Dataset-ISQ) | ≥50% | Per-vessel (n=1,956) | 89.2 (86.3, 91.8) | 97.1 (96.2, 98.0) | 90.7 (87.9, 93.5) | 96.6 (95.6, 97.4) |
| Per-patient (n=652) | 91.9 (88.3, 94.9) | 93.2 (90.7, 95.7) | 90.6 (87.1, 93.9) | 94.1 (91.6, 96.2) | ||
| ≥70% | Per-vessel (n=1,956) | 89.8 (84.9, 94.1) | 98.4 (97.9, 98.9) | 83.4 (77.6, 88.5) | 99.1 (98.6, 99.5) | |
| Per-patient (n=652) | 94.2 (89.2, 98.1) | 95.8 (94.1, 97.4) | 80.8 (73.5, 87.7) | 98.9 (97.9, 99.6) | ||
| External (Dataset-ESQ) | ≥50% | Per-vessel (n=768) | 89.0 (82.7, 93.8) | 97.3 (96.0, 98.6) | 87.7 (81.4, 93.3) | 97.6 (96.3, 98.7) |
| Per-patient (n=256) | 88.0 (81.0, 94.4) | 94.5 (90.7, 97.6) | 90.0 (83.3, 95.6) | 93.4 (89.2, 96.8) | ||
| ≥70% | Per-vessel (n=768) | 86.8 (76.8, 95.3) | 98.6 (97.6, 99.4) | 82.1 (69.8, 92.3) | 99.0 (98.3, 99.7) | |
| Per-patient (n=256) | 91.9 (82.6, 100.0) | 97.3 (94.9, 99.1) | 85.0 (72.5, 94.6) | 98.6 (96.8, 100.0) |
Values in parentheses are 95% confidence intervals. AI, artificial intelligence; PPV, positive predictive value; NPV, negative predictive value; ISQ, internal stenosis quantification; ESQ, external stenosis quantification.