Skip to main content
. 2023 Jun 15;37(18):3813–3818. doi: 10.1038/s41433-023-02622-9

Table 4.

Evaluation of the generalization capability of the algorithm on images with/without the features of some retinal comorbidities such as DR and AMD on validation dataset 1 and validation dataset 2.

Sensitivity (95% CI) Specificity (95% CI) AUC (95% CI) P-value
Set 1 Set 2 Set 1 Set 2 Set 1 Set 2 Set 1 Set 2
With Comorbidities 1.000 (1.000–1.000) 0.922 (0.916–0.928) 0.825 (0.811–0.838) 0.939 (0.933–0.944) 0.970 (0.954–0.987) 0.975 (0.964–0.985) 4.55e–11 1.97e–6
Without Comorbidities 0.893 (0.882–0.903) 0.912 (0.905–0.918) 0.944 (0.935–0.952) 0.913 (0.906–0.919) 0.943 (0.922–0.965) 0.963 (0.954–0.972)
Overall Performance 0.918 (0.908–0.927) 0.914 (0.907–0.920) 0.934 (0.925–0.942) 0.922 (0.916–0.928) 0.952 (0.945–0.968) 0.966 (0.959–0.974)

DR diabetic retinopathy, AMD age-related macular degeneration, AUC area under the curve, CI confidence interval, Set 1: validation dataset 1, Set 2: validation dataset 2.