Use of a Post-Acquisition Centering Method Can Improve the Repeatability of Thickness Measurements in SD-OCT Volumes

Abstract

 

Purpose:

 

In a clinical setting, acquisitions of SD-OCT volumetric scans may not be consistently centered across visits, thus potentially affecting the repeatability of resulting thickness measurements. Here, we compared the repeatability of a 3-D automated layer segmentation approach for measuring regional RNFL thickness with and without the use of a post-acquisition centering method.

 

Methods:

 

ONH-centered and macula-centered Cirrus SD-OCT volumes were acquired for both eyes of 10 subjects with glaucoma on each of two visits. The time between visits was less than 3 months. The RNFL layer was segmented on each volume using our previously published 3-D automated graph-theoretic approach. In the ONH-centered scans, the mean thickness was assessed in 12 clock sectors within 5 radial bands around the center, resulting in 60 measurements. For the macula scans, we also included the central circular region, resulting in a total of 61 measurements. The repeatability of the RNFL thickness was assessed within these sectors without centering the scans. The repeatability was then re-assessed after centering the scans using a simple manual registration method.

 

Results:

 

In the ONH scans, a significant difference was noted between the mean unsigned difference computed before and after centering: 14.46 ± 6.50 µm and 9.19 ± 4.38 µm (p < 0.001), which is depicted in the figure. A similar significant difference was noted in the macula scans, where the mean unsigned difference reduced from 3.47 ± 1.09 µm to 3.04 ± 0.99 µm (p < 0.001) after centering the scans.

 

Conclusions:

 

The use of a post-acquisition centering method can improve the repeatability of RNFL thickness measurements from OCT scans. Further tests using more robust centering techniques could provide more reliable layer thickness assessments.  

 

Keywords: image processing • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: non-clinical