A Comparison of Methods for Correcting Ocular Magnification to Reduce the Variance of Normal Retinal Nerve Fiber Layer Thickness Measurements

Abstract

 

Purpose

 

To compare the effects on retinal nerve fiber layer (RNFL) thickness measurements of different ocular magnification correction methods used to scale macular thickness profiles from frequency domain optical coherence tomography (fdOCT) scans.

 

Methods

 

Macular horizontal line fdOCT scans (Spectralis, Heidelberg, Inc.) were obtained from 155 healthy control eyes. An RNFL thickness profile (thickness as a function of distance from the fovea) was calculated for each scan by hand-correction of a previously validated automated segmentation algorithm.[1,2] Using biometric data, including axial length (AL, 24.18 ± 2.58 mm), corneal curvature (CC, 7.76 ± 0.54 mm), and spherical equivalent of refractive error (RE, -1.12 ± 4.73 D), scan lengths were adjusted by horizontally scaling profiles according to each of 4 different methods: an automatic partial correction (APC) made by the fdOCT manufacturer’s software, based on scan focus (SF, similar to RE); full correction (FC) by the manufacturer’s software, based on SF and manually-inputted CC values; Littman’s original correction (LC), based on RE and CC; and Bennett’s abbreviated correction (BC), based on AL alone.[3] Mean RNFL thickness (MRT) over the region from the foveal center to 3 mm nasal was calculated. F-tests were used to compare variances in MRTs across individuals.

 

Results

 

All methods significantly reduced the variance of MRTs (p < 0.01). Uncorrected MRTs (mean = 30.8 μm) had a standard deviation (SD) of 3.95 μm (Fig. 1A), whereas for corrected MRTs, SDs were 3.20 μm (APC), 3.19 μm (FC), 3.01 μm (LC, Fig. 1B), and 3.18 μm (BC). Variances of MRTs from correction via manufacturer methods (APC & FC, p = 0.98), and from correction via Littman-derived methods (LC & BC, p = 0.50), were not statistically different.

 

Conclusions

 

Correcting for ocular magnification reduced the variability in RNFL thickness measurements among healthy controls, although the choice of method did not significantly affect the results. Thus it is reasonable to choose the most convenient correction method based upon the biometric data and the OCT machine available. 1. Raza et al. AO, 2011; 2. Yang et al Biomed Opt Exp, 2011; 3. Bennett et al. Graefe’s, 1994.

 

 

Figure 1. Individual RNFL thickness profiles (light blue) with mean ± 2 SD (dark blue). (A) Uncorrected profiles. (B) Profiles corrected by LC.

 

Keywords: 552 imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • 630 optical properties • 610 nerve fiber layer