Impact of segmentation errors and retinal blood vessels on retinal nerve fibre layer measurements using spectral-domain optical coherence tomography

Abstract

Purpose To investigate the impact of retinal blood vessels and retinal nerve fibre layer (RNFL) segmentation errors on RNFL measurement. Methods One eye of 180 subjects (60 normal, 66 mild-to-moderate and 54 advanced glaucoma subjects) was randomly selected for RNFL imaging with a spectral-domain OCT. The boundaries of the RNFL detected by the instrument software were checked, and the segmentation errors were corrected by a customized computer program. The differences in average and regional RNFL thicknesses (RNFLT) before and after the correction were analysed to determine the frequency of segmentation error (defined as an absolute difference in average RNFLT >5.0 μm). The ratio of retinal blood vessel cross-sectional area to RNFL cross-sectional area was calculated. Results The difference in average RNFLT (postsegmentation minus presegmentation refinement) ranged from -3.0 to 2.5 μm (mean ± standard deviation: 0.83 ± 0.86 μm) in the normal, -2.5 to 5.0 μm (0.56 ± 1.08 μm) in the mild-to-moderate glaucoma and -11.0 to 9.5 μm (0.05 ± 3.49 μm) in the advanced glaucoma groups (p = 0.003). A total of 15% of eyes had average RNFLT measurement error >5.0 μm in the advanced glaucoma group. The proportion of retinal blood vessels in the RNFL also increased with the severity of glaucoma (p < 0.001) with 4.2 ± 1.0% in the normal group, 4.9 ± 1.5% in the mild-to-moderate and 8.5 ± 3.5% in the advanced glaucoma groups. Conclusions Inclusion of retinal blood vessels and RNFL segmentation error could adversely affect RNFL measurement, particularly in advanced glaucoma when the RNFL was thin.