Reproducibility of automated choroidal thickness measurements: swept-source OCT and spectral-domain OCT using enhanced depth imaging

Abstract

 

Purpose

 

The purpose of this study was to evaluate the reproducibility of a fully automated three-dimensional method capable of segmenting the entire choroid and quantifying choroidal thicknesses in optical coherence tomography (OCT) image data from two different deep-range imaging techniques, swept-source (SS)-OCT using a 1050nm wavelength and spectral-domain (SD)-OCT using enhanced depth imaging (EDI).

 

Methods

 

Nine healthy subjects underwent OCT imaging using two different machines from Topcon (DRI OCT-1) (512×256×992 voxels, 6.0×6.0×2.6mm³, voxel size of 11.7×23.4×2.6μm³) and Spectralis (512×97×496 voxels, 5.9×5.9×1.9mm³, voxel size of 11.4×60.9×3.9μm³) using EDI within one hour. Nine independent macula-centered volumetric scans were obtained from each scanner (18 in total). First, a shape-prior based soft-constraint graph-search method was employed to segment Bruch’s membrane (BM), which is modeled as a convex arc on each B-scan. Second, a combined graph-cut-graph-search method was utilized to segment the posterior choroidal surface. Choroidal thickness maps were derived from local distances between the posterior choroidal surface and Bruch’s membrane. To evaluate the test-retest reproducibility between two SD-OCTs, coefficients of variation (CV) of the average choroidal thickness were determined using the root mean square (RMS) approach. Differences of the average choroidal thickness were also reported.

 

Results

 

The results showed a high correlation between the two evaluated OCT scanners (R2 = 0.97). The RMS CV was 6.8%. The difference assessment showed an average choroidal thickness bias of 24.7µm (8.1% of the total choroidal thickness) for the two OCT scanners.

 

Conclusions

 

We have developed a fully automated three-dimensional method for segmenting the choroidal layer and quantifying choroidal thickness. The method yielded highly robust segmentation results on Topcon SS-OCT and Spectralis SD-OCT scans. Our results have a potential to facilitate multi-center studies with heterogeneous device utilization.

 

 

Figure 1. Segmentation results and choroidal thickness map for Topcon SS-OCT scanner (left) and Spectralis SD-OCT scanner (right).

 

 

Figure 2. (a) Cross-scanner reproducibility assessment. (b) Assessment of cross-scanner choroidal thickness differences.

 

Keywords: 549 image processing • 550 imaging/image analysis: clinical • 452 choroid