Table 3.
Comparison of QT prolongation identified by NLP in ECG impressions to automated QTc by ECG machine
| QT prolongation queries |
ECG machine calculations |
|||||
|---|---|---|---|---|---|---|
| Regular expression | NLP | QTc >400 | QTc >450 | QTc >500 | QTc >550 | |
| ECGs matching criteria | 2,413 | 2,364 | 35813 | 11,630 | 2,260 | 428 |
| True positives | 2,370 | 2,364 | 2358 | 2301 | 507 | 108 |
| False negatives | 3 | 9 | 15 | 72 | 1866 | 2265 |
| False positives | 43 | 0 | 33,455 | 9,329 | 1,753 | 320 |
| True negatives | 41,908 | 41,945 | 8,490 | 32,616 | 40,192 | 41,625 |
|
| ||||||
| Sensitivity | 0.999 | 0.996 | 0.994 | 0.970 | 0.214 | 0.046 |
| Specificity | 1.000 | 1.000 | 0.202 | 0.778 | 0.958 | 0.992 |
| Positive predictive value | 0.982 | 1.000 | 0.066 | 0.198 | 0.224 | 0.252 |
| Negative predictive value | 1.000 | 1.000 | 0.998 | 0.998 | 0.956 | 0.948 |
|
| ||||||
| F-measure | 0.990 | 0.998 | 0.124 | 0.329 | 0.219 | 0.077 |