Table 1

Performance of the retinopathy of prematurity deep learning (i-ROP DL) system for detecting various parameters and levels of ROP

Frequency n (%)†AUROC (95% CI)
Diagnostic p arameter*
Plus disease128 (3)0.989 (0.984 to 0.994)
Stage 3299 (6)0.880 (0.860 to 0.899)
 Stage 3 without pre-plus or plus68 (1)0.672 (0.628 to 0.716)
Zone I222 (5)0.817 (0.786 to 0.849)
 Zone I without pre-plus or plus90 (2)0.616 (0.565 to 0.667)
Disease c ategory*
Clinically significant ROP912 (19)0.914 (0.903 to 0.925)
 Type 1 ROP155 (3)0.960 (0.941 to 0.978)
 Type 2 ROP300 (6)0.867 (0.854 to 0.880)
 Pre-plus disease636 (13)0.910 (0.900 to 0.920)
  • Results are displayed as area under the receiver operating curve (AUROC) compared with a reference standard diagnosis.

  • *According to reference standard diagnosis.

  • †Out of total of 4861 eye examinations from 870 infants.