A metrological approach to the analysis of choroidal thickness by OCT in the context of myopia research

Abstract

Purpose : Studies found the choroid to be able to change its thickness in response to defocus in a bi-directional fashion. However, measuring choroidal thickness from optical coherence tomography (OCT) images can cause difficulties due to a limited visibility of the choroidal-scleral interface. The study evaluated the agreement of a freely available software for automated choroidal segmentation in two spectral-domain OCT compared to manual segmentation.

Methods : Three consecutive OCT volume scans were performed in 23 healthy subjects using two different spectral-domain OCT (Cirrus HD-OCT 5000, Carl Zeiss Meditec AG, Dublin, CA, USA and HRA+OCT Spectralis, Heidelberg Engineering, Heidelberg, Germany). Subfoveal choroidal thickness was obtained by manual measurements of one naïve (Examiner 1) and one experienced examiner (Examiner 2), additionally to an automated segmentation software developed by Mazzaferri et al. (2017). Volume scans were chosen as scan pattern, as the automated software is able to analyze choroidal thickness in OCT volume scans in particular. Subjects with a difference of ≥100µm in choroidal thickness between segmentation methods were excluded. ICC, 95% Limits of Agreement (LoA) and Coefficients of Repeatability (CoR) were calculated.

Results : The repeatability of three measurements was between 18µm and 26µm for different segmentation methods and examiners, whereas automated segmentation yielded less repeatable results compared to manual segmentation. The mean difference between methods ranged from +3µm to +15µm, indicating higher thickness values with the Cirrus OCT for all segmentation methods. More detailed results are reported Table 1. Figure 1 shows the Bland-Altman plot for the results from the automated software and both examiners in the two OCT devices.

Conclusions : Repeatability and agreement values exceed previously reported effect sizes of choroidal thickness changes in response to defocus. The obtained metrological results should be considered for the future evaluation and interpretation of changes in choroidal thickness in myopia research.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.

 

Table 1: Agreement and repeatability analysis of the different segmentation methods for both devices. MD = Mean difference, LoA = Limit of Agreement, ICC = Intra-class correlation coefficient, CI = Confidence interval, CoR = Coefficient of Repeatability.

 

Figure 1: Inter-device Bland-Altman analysis for all three segmentation methods.