Table 2.
Performances of artificial intelligence models for the detection of sacroiliitis compatible with the assessment of spondyloArthritis international society definition.
| Performances by individual MRI slices | |||
|---|---|---|---|
| Sensitivity (95% CI) | Specificity (95% CI) | AUROC (95% CI) | |
| Method A a | 0.517 (0.493–0.780) | 0.944 (0.933–0.955) | 0.731 (0.681–0.780) |
| Method B b | 0.563 (0.538–0.587) | 0.933 (0.920–0.945) | 0.747 (0.699–0.796) |
| Method C c | 0.725 (0.705–0.745) | 0.936 (0.924–0.947) | 0.830 (0.792–0.868) |
| Performances by each patient | |||
| Sensitivity | Specificity | AUROC | |
| Method A a | 0.806 (0.729–0.883) | 0.617 (0.523–0.711) | 0.711 (0.660–0.763) |
| Method B b | 0.859 (0.792–0.926) | 0.589 (0.494–0.684) | 0.722 (0.671–0.774) |
| Method C c | 0.947 (0.912–0.982) | 0.691 (0.603–0.779) | 0.816 (0.776–0.856) |
AUROC, area under the receiver operating characteristic curve.
a
Artificial intelligence model for the detection of sacroiliitis without augmentation and maximum intensity projection.
b
Artificial intelligence model for the detection of sacroiliitis using augmentation without maximum intensity projection.
c
Artificial intelligence model for the detection of sacroiliitis using both augmentation and maximum intensity projection.