Table 2.
Performance of AI-stenosis (CathAI) versus REPORT-stenosis in the test dataset.
| Number of REPORT-stenosis labels | AUC (95% CI) to discriminate </≥ 70% stenosis severity | Mean Absolute difference, % stenosis (continuous) ± SD | r* (95% CI) | ICC* (95% CI) | Sensitivity (95% CI)** | Specificity (95% CI)** for </≥ 70% stenosis severity | PPV (95% CI)** for </≥ 70% stenosis severity | NPV (95% CI)** for </≥ 70% stenosis severity | P value for Interaction | |
|---|---|---|---|---|---|---|---|---|---|---|
| Overall (Artery level) | 1734 | 0.862 (0.843–0.880) | 17.9 ± 15.5 | 0.74 (0.72–0.76) | 0.72 (0.60–0.84) | 74.5 (70.0–78.4) | 78.1 (76.1–80.1) | 46.1 (42.2–50.2) | 92.4 (90.9–93.6) | – |
| Left Coronary Artery | 980 | 0.835 (0.808–0.862) | 19.0 ± 15.8 | 0.72 (0.68–0.74) | 0.69 (0.54–0.85) | 76.1 (71.0–81.3) | 72.2 (68.9–75.3) | 42.6 (38.0–47.1) | 93.3 (91.8–95.0) | Ref |
| Right Coronary Artery | 754 | 0.894 (0.869–0.919) | 16.4 ± 15.0 | 0.77 (0.74–0.80) | 0.75 (0.65–0.87) | 70.9 (62.7–78.0) | 85.6 (83.2–87.9) | 50.1 (43.0–57.1) | 82.7 (80.1–85.6) | 0.11 |
| Sex*** | ||||||||||
| Male | 1058 | 0.830 (0.805–0.856) | 18.5 ± 15.3 | 0.73 (0.70–0.75) | 0.68 (0.55–0.83) | 78.6 (73.8–82.9) | 70.6 (68.1–73.3) | 47.7 (43.5–52.5) | 90.6 (88.2–92.5) | Ref |
| Female | 672 | 0.890 (0.864–0.923) | 16.9 ± 15.8 | 0.75 (0.63–0.87) | 0.74 (0.70–0.77) | 80.2 (72.0–88.4) | 78.9 (75.6–82.0) | 35.6 (29.2–40.2) | 96.6 (95.2–97.9) | 0.02 |
ICC intra-class correlation, r Pearson correlation, AUC area under the receiver operating characteristic curve, PPV positive predictive value, NPV negative predictive value, MSE mean squared error.
I*Calculation is between AI-stenosis (continuous) and REPORT-stenosis; **The threshold for determining sensitivity, specificity, PPV and NPV for AI-stenosis was 54%. The interaction P values were calculated using two-sided Wald tests between the CathAI Algorithm 4 probability and the respective covariates for stenosis percentage. P values < 0.05 are bolded. ***Sex for n = 4 patients was unknown.