Exploratory structure-function associations in retinal vein occlusion-associated macular edema using deep learning quantification of OCT biomarkers and microperimetry

OBJECTIVE: To investigate associations between disease-specific OCT biomarkers and pointwise retinal sensitivity (PWS) in retinal vein occlusion (RVO) derived from 2 microperimetry (MP) devices. DESIGN: Exploratory, prospective, cross-sectional study. PARTICIPANTS: Twenty eyes of 20 patients with macular edema secondary to RVO. METHODS: Patients underwent spectral-domain OCT (Spectralis; Heidelberg Engineering) and 2 examinations with each MP device (MAIA2, iCare; MP-3, Nidek). Intraretinal fluid (IRF), subretinal fluid (SRF), and ellipsoid-zone thickness (EZT) were segmented using deep-learning-based algorithms; disorganisation of inner retinal layers (DRIL) and intraretinal hemorrhage were annotated manually. Co-registration of OCT volumes and MP grids enabled pointwise structure-function analyses. Mixed-effects models assessed the associations of biomarkers with PWS, pointwise EZT, and fixation stability (FS). RESULTS: Mean age was 66 ± 13 years, 13 patients were male, and 11 eyes had branch and 9 central RVO. Greater EZT was associated with greater PWS (+0.14 dB/µm), whereas greater IRF volume (-1.15 dB/nL) and the presence of SRF (-2.35 dB), DRIL (-3.29 dB), and hemorrhage (-9.33 dB) were associated with lower PWS (all P < 0.001). Greater IRF volume (-0.46 µm/nL) and SRF presence (-6.76 µm) were associated with lower pointwise EZT (both P < 0.001). CONCLUSIONS: IRF, SRF, hemorrhage, DRIL, and EZT showed associations with PWS, whereas IRF and SRF were associated with lower EZT in RVO. MP may improve understanding of structure-function associations in RVO.