Purchase this article with an account.
Cong Ye, Dennis S. Lam, Christopher K. Leung; Investigation of Floor Effect for OCT RNFL Measurement. Invest. Ophthalmol. Vis. Sci. 2011;52(14):176.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
OCT average RNFLT does not get below 30µm even in eyes with optic neuropathy with no light perception. However, it is unknown what factors contribute to this "floor effect". We examined the effect of retinal blood vessels (BV) and the frequency of segmentation failure on RNFL measurement obtained with a spectral-domain OCT.
130 glaucoma, 40 suspects and 40 normal subjects were enrolled. One eye was selected at random from each subject for circumpapillary RNFL scan by the Spectralis OCT. The OCT images were exported to a customized computer program written in MATLAB to remove BVs and refine the boundaries of RNFL. The effect of BVs removal and RNFL segmentation refinement to RNFL measurement were evaluated.
The mean proportion of BVs relative to the average RNFLT was 4.2±0.9%, 4.7±1.5% and 7.8±4.7%, respectively for the normal, suspect and glaucoma groups. When the average RNFLT was above approximately 70µm, the proportion remained at 4.5% (95% CI: 4.2 - 4.8%). When it is below 70µm, the proportion increased with decreasing RNFLT. Taking the refined RNFLT as the reference standard, 7.50% had an underestimate (average RNFLT (post-refinement) - average RNFLT (pre-refinement) > 3.5µm) in the suspect group, 6.15% had an overestimate (average RNFLT (pre-refinement) - average RNFLT (post-refinement) > 3.5µm), and 4.62% had an underestimate in the glaucoma group. For eyes with an overestimate, the algorithm misidentified the outer boundary of the ganglion cell layer as that of the RNFL (Fig. 1A). For eyes with an underestimate, the algorithm failed to detect the segment with thin RNFL (Fig. 1B).
Both the inclusion of BVs and segmentation failure could contribute to floor effect of OCT RNFL measurement. The impact of BVs on RNFL measurement is more substantive in advanced glaucoma when the RNFL is thin. Enhancement of RNFL segmentation algorithm with removal of BVs may improve the detection of progressive RNFL changes.
This PDF is available to Subscribers Only