Table 3.
Confusion matrices
| All participants | LVEF ≤40% | LVEF >40% | |
|---|---|---|---|
| Position 2 | |||
| Number | 979 | 99 | 880 |
| AI-ECG positive | 352 | 84 | 268 |
| AI-ECG negative | 627 | 15 | 612 |
| Positions 2 and 5 (rule based) | |||
| Number | 864 | 81 | 783 |
| AI-ECG positive | 224 | 67 | 157 |
| AI-ECG negative | 640 | 14 | 626 |
| Positions 2 and 5 (logistic regression) | |||
| Number | 346 | n=37 | 309 |
| AI-ECG positive | 95 | 34 | 61 |
| AI-ECG negative | 251 | 3 | 248 |
Confusion matrices are displayed according to the restricted threshold for maximising sensitivity and specificity, with rule sensitivity >81, specificity >67; or sensitivity >81, maximising specificity. AI=artificial intelligence. LVEF=left ventricular ejection fraction.