Table 1.
Comparison of diagnostic performances of the AI-based software for subsets (A)–(C).
| Sensitivity | p-value | Specificity | p-value | PPV | p-value | NPV | p-value | Accuracy | p-value | |
|---|---|---|---|---|---|---|---|---|---|---|
| (A) Including all lesions | 67.2 (62.2–72.1) | – | 91.1 (89.9–92.4) | – | 57.7 (52.9–62.5) | – | 93.9 (92.8–95) | – | 87.5 (86.1–88.8) | – |
| (B) Excluding cardiomegaly | 65.4 (60.4–70.4) | 0.014¶ | 93.9 (92.8–94.9) | < 0.001¶ | 65.8 (60.8–70.8) | < 0.001¶ | 93.8 (92.7–94.9) | 0.310¶ | 89.5 (88.3–90.8) | < 0.001¶ |
| (C) Excluding cardiomegaly and children ≤ 2 years old | 86.4 (80.5–92.2) | < 0.001¥ | 97.9 (97.1–98.7) | < 0.001¥ | 79.7 (73.1–86.3) | < 0.001¥ | 98.7 (98.1–99.3) | < 0.001¥ | 96.9 (96.0–97.8) | < 0.001¥ |
Values are presented in percentages (%) with 95% confidence intervals.
¶Comparison between subset (A) and (B). ¥Comparison between subset (B) and (C).
CI confidence interval, PPV positive predictive value, NPV negative predictive value.