Related Experiment Video
Updated: Aug 6, 2026
An Acute Retinal Model for Evaluating Blood Retinal Barrier Breach and Potential Drugs for Treatment
Published on: September 13, 2016
Robustness of automated Bruch's membrane opening minimum rim width and retinal nerve fiber layer thickness without
Valeria Batista Boreck Seki1, George Cunha Cardoso2, Renato Antunes Schiave Germano3
1Departamento de Oftalmologia, Otorrinolaringologia e Cirurgia de Cabeça e Pescoço, Faculdade de Medicina de Ribeirão Preto, Universidade de São Paulo, Av Bandeirantes, 3900 - Monte Alegre, Ribeirão Preto, São Paulo, CEP: 14.049-900, Brazil.
Purpose:
To evaluate the structure-structure correspondence between circumpapillary retinal nerve fiber layer (cpRNFL) thickness and Bruch's Membrane Opening Minimum Rim Width (BMO-MRW) in primary open-angle glaucoma (POAG) and controls, and to assess the impact of manual realignment to the clinically visible optic nerve head (ONH) margin.
Methods:
In this cross-sectional study, spectral-domain optical coherence tomography measured BMO-MRW at 48 radial positions and the 3.5-mm cpRNFL thickness. The Spectralis normative database was used to confirm sectoral MRW and cpRNFL correlation in a large, independent dataset within Garway-Heath sectors. An internal anatomical reference curve of MRW-cpRNFL differences was constructed from 36 eyes with concordant automated and clinically adjusted ONH margins. Subsequently, 116 eyes (72 control, 44 POAG) with margin discordance were analyzed to determine whether manual realignment by two examiners improved correspondence with the reference curve, using squared Spearman correlation (rho²).
Results:
Automated BMO-MRW and cpRNFL showed significant sectoral correlation (maximum r = 0.44, inferotemporal; p < 0.001), with excellent interobserver agreement (ICC 0.99 controls, 0.96 POAG). In controls, automated curves were closer to the reference than manual realignment (9.9 ± 9.8 μm vs. 27.7 ± 16.1 μm; p < 0.0001; rho²=0.96 vs. 0.91). In POAG, manual realignment reduced mean deviation from the reference (- 99.7 ± 37.7 μm vs. - 124.3 ± 30.4 μm; p < 0.0001). Nevertheless, automated segmentation retained stronger monotonic association (rho²=0.82 vs. 0.57), indicating better structural correspondence. Realignment offered minimal benefit in inferotemporal sectors.
Conclusions:
Automated BMO-based segmentation demonstrates consistent structure-structure correspondence with cpRNFL in healthy and glaucomatous eyes. Manual realignment yields limited benefit in relatively preserved rim sectors and adds little value in routine glaucoma assessment.
