Correction of the Retinal Angle in SD-OCT of Glaucomatous Eyes Provides Better Reproducibility of Optic Nerve Head Parameters

Abstract

 

Purpose

 

We report a novel method for correction of the retinal angle in the spectral-domain optical coherence tomography (SD-OCT) of glaucomatous eyes. The performed study compares the reproducibility of optic nerve head (ONH) parameters obtained by our custom algorithms without/with the correction step.

 

Methods

 

Repeat ONH-centered OCT volumes (200 × 200 × 1024 voxels, 6 × 6 × 2 mm3) were obtained within a 4-month period from both eyes of 48 glaucoma patients and suspects (62.9 ± 12.5 years, 20 males) using a Cirrus machine. Correction of the retinal angle in the OCT volumes starts by fitting a spline to the outer boundary of the retinal pigment epithelium (RPE) while excluding the ONH region for each B-scan in the data domain (Fig. 1B). The predominant direction of the depicted retina is derived from the spline (Fig. 1C) and the angle (θ1) is calculated by employing the equation in Fig. 1. The x- and z-directional voxel sizes (30.15 µm, 1.96 µm) and the angle (θ2) of the line in the physical domain (Fig. 1E) are used for calculation of θ1. The B-scan rotated by θ1 (Fig. 1E) is straightened by re-arranging rotated A-scans (Fig. 1F). With/without retinal-angle correction, retinal nerve fiber layers (RNFL) and optic discs were automatically segmented by our previously-reported graph-theoretic approach and voxel-column classification method, respectively.

 

Results

 

The reproducibility of the mean circular RNFL thicknesses, optic disc areas, neuroretinal rim areas, optic cup areas, linear cup-to-disc radios (LCDRs) obtained from the repeat OCT volumes was determined by intraclass correlation coefficient (ICC), coefficient of variation (CV), and repeatability coefficient (RC) provided in Table 1 for our methods without/with the retinal angle correction step.

 

Conclusions

 

When analyzing repeat OCT image volumes, the ONH parameters measured in a consistent retinal shape after the correction of the retinal angle showed better reproducibility of RNFL thickness, disc area, rim area, and cup area when compared with reproducibility obtained from uncorrected OCT volumes (p < 0.05).

 

 

Figure 1. (A) B-scan in the data domain. (B) Spline fitted to the outer boundary of the RPE excluding the ONH region. (C) Retinal direction derived from the spline. (D) B-scan in the physical domain. (E) B-scan rotated clockwise by θ1. (F) Straightened B-Scan.

 

 

Table 1. Reproducibility of ONH parameters.

 

Keywords: 549 image processing