Automated OCT-tailored and same-visit biomarker-targeted microperimetry in geographic atrophy

This proof-of-concept study developed and validated an automated, biomarker-targeted, real-time microperimetry (MP) approach for individualized structure-function assessment in geographic atrophy (GA). Patients underwent volumetric spectral-domain optical coherence tomography (OCT) and mesopic MP. Automated deep learning-based OCT segmentation using a clinically validated software was applied to identify regions of ellipsoid zone loss (EZL) and retinal pigment epithelium loss (RPEL). A rule-based algorithm then selected 40 biomarker-targeted retinal test locations, with test-point density dynamically adapted to individual lesion morphology and deliberately prioritized within EZL rather than uniformly distributed. Statistical analysis employed multivariable mixed-effects models. In 64 eyes from 44 patients (mean age 79.1 ± 5.0 years; 59% female), 2,560 targeted MP stimuli were tested. Mean retinal sensitivity (RS) was 7.2 ± 7.4 dB in RPEL, 13.3 ± 7.1 dB in EZL, and 19.0 ± 5.0 dB in preserved retina (p < 0.001). The median deviation between intended and actual test locations was 38.2 μm (IQR 24.6–59.3 μm), with 98.5% of points correctly localized. This OCT-based, biomarker-guided MP approach enables precise, patient-specific structure-function mapping within a single visit and may serve as a sensitive functional endpoint in future GA trials and clinical practice.