Beyond the Index: Association of AI-CO Central Image Area with Central Visual Field Damage in Glaucoma
In the authors' words
Purpose: To evaluate the associations of the numerical AI-CO center index and visually classified central dark-area extent with Humphrey 10-2 mean deviation (MD). Methods: This single-center retrospective study included 320 eyes of 169 patients with AI-CO imaging, broad-field 24-2F or 30-2 testing, and 10-2 testing. Both visual field examinations were performed within 365 days before or after AI-CO imaging. Two ophthalmologists independently classified central dark-area extent into five Levels (0-4). Analyses assessed Spearman correlations, areas under the receiver operating characteristic curve (AUCs), distributions across final central dark-area Levels, and interrater reliability. Patient-cluster bootstrap confidence intervals and a one-eye-per-patient sensitivity analysis addressed inclusion of both eyes. Results: AI-CO center correlated with 10-2 MD (rho=0.829; 95% CI, 0.773-0.873). AUCs for 10-2 MD below -2, -6, and -12 dB were 0.902, 0.939, and 0.961, respectively. The numbers of eyes in Levels 0-4 were 66, 57, 94, 53, and 50; median 10-2 MD values were -0.59, -1.48, -3.54, -6.24, and -18.15 dB, respectively. Quadratic-weighted kappa was 0.970 (95% CI, 0.958-0.981). Sensitivity analyses yielded broadly similar findings. Conclusions: The numerical AI-CO center index and visually classified central dark-area extent were associated with central visual field damage. Assessing the extent of dark regions in the central image alongside the numerical index may provide useful information for interpreting central visual field status from the AI-CO display.
Appeared: Thursday, October 22. medRxiv. Preprint, not yet peer-reviewed.