Table 3. Performance characteristics of the classifier.
| Characteristic | TP | FP | TN | FN | Sn, % (95%CI) | Sp, % (95%CI) |
| Development dataset, reports n = 366 | 197 | 32 | 117 | 20 | 91 (86 to 94) | 79 (71 to 84) |
| 1Held-out dataset, reports n = 83 | 35 | 13 | 30 | 5 | 88 (74 to 95) | 70 (55 to 81) |
| All reports, n = 449 | 232 | 45 | 147 | 25 | 90 (86 to 93) | 77 (70 to 82) |
1
Held out dataset were annotated at report level only as being positive, negative or equivocal for IMD.
Abbreviations: TN, true positives; FP, false positives; TN, true negatives; FN, false negatives; Sn, sensitivity; Sp, specificity; CI, confidence interval.