Effect of image processing on sensitivity and specificity for detection of optic neuritis on FLAIR and contrast-enhanced T1WI by 6 masked readers including 2 radiology residents (readers 1, 2), 2 neuroradiology fellows (readers 3, 4), and 2 attending neuroradiologists (readers 5, 6)
| Reader | 1 | 2 | 3 | 4 | 5 | 6 | Average |
|---|---|---|---|---|---|---|---|
| Sensitivity for baseline FLAIR image | 18/28 | 16/28 | 19/28 | 27/28 | 19/28 | 21/28 | 71.4 ± 13.6 |
| 64% | 57% | 68% | 96% | 68% | 75% | ||
| Sensitivity for processed FLAIR image | 26/28 | 26/28 | 26/28 | 27/28 | 27/28 | 27/28 | 94.6 ± 2.0 |
| 93% | 93% | 93% | 96% | 96% | 96% | ||
| Specificity for baseline FLAIR image | 28/28 | 28/28 | 26/28 | 20/28 | 26/28 | 27/28 | 92.3 ± 10.7 |
| 100% | 100% | 93% | 71% | 93% | 96% | ||
| Specificity for processed FLAIR image | 26/28 | 25/28 | 26/28 | 24/28 | 26/28 | 27/28 | 91.7 ± 3.7 |
| 93% | 89% | 93% | 86% | 93% | 96% | ||
| P valuea | <.001 | <.001 | .02 | .02 | .01 | .01 | |
| Sensitivity for baseline contrast-enhanced image | 17/28 | 18/28 | 20/28 | 25/28 | 19/28 | 22/28 | 72.0 ± 10.5 |
| 61% | 64% | 71% | 89% | 68% | 79% | ||
| Sensitivity for processed contrast-enhanced image | 22/28 | 20/28 | 21/28 | 23/28 | 24/28 | 22/28 | 78.6 ± 5.1 |
| 79% | 71% | 75% | 82% | 86% | 79% | ||
| Specificity for baseline contrast-enhanced image | 18/18 | 18/18 | 18/18 | 13/18 | 17/18 | 18/18 | 94.4 ± 11.1 |
| 100% | 100% | 100% | 72% | 94% | 100% | ||
| Specificity for processed contrast-enhanced image | 18/18 | 18/18 | 18/18 | 17/18 | 17/18 | 18/18 | 98.1 ± 2.9 |
| 100% | 100% | 100% | 94% | 94% | 100% | ||
| P valuea | <.001 | <.001 | .002 | .12 | .01 | .003 |
↵a P value for improvement of diagnostic performance in the optic neuritis eyes compared with the control eyes.