Purpose
To compare retinal sublayer thickness and intensity profiles from two spectral-domain optical coherence tomography (SD-OCT) devices.
Methods
Twenty Cirrus and Spectralis SD-OCT images from 20 eyes of 20 normal subjects were used in this IRB-approved study. Scan acquisition protocols were based on that typical of most clinical trials and practice at the University of Southern California. Each volume scan consisted of a macular cube of 1024 * 512 * 128 voxels for Cirrus (no averaging) and 496 * 1024 * 37 voxels for Spectralis (9x averaging). The physical size for each Cirrus volume was 2mm * 6mm * 6mm. However, it varied slightly for each Spectralis volume in a machine-provided average 1.92mm * 5.91mm * 4.58mm. Eleven retinal surfaces were segmented using a graph-based algorithm and manually corrected by a certified reading center grader (AH) and reviewed by the Principle Investigator (SRS). The mean thickness and intensity profiles were computed, plotted, and compared.
Results
Figure 1 illustrates the mean thickness and intensity profiles for all the segmented layers for the two devices. When comparing between devices, a paired t-test shows no statistically significant difference (p > 0.05) between the retinal sublayers and choroid, with a mean absolute difference 2.85 ± 6.10µm for the retinal sublayers and 20.81µm for the choroid. In contrast, the mean intensity for all of the segmented layers (including visible vitreous as well) was statistically significantly different (p < 0.01), with a mean absolute difference 39 ± 31 intensity units.
Conclusions
Automated retinal sublayer thicknesses were similar between the two OCT devices, despite differences in the relative intensity/brightness of the various layers. The differences in relative brightness between layers likely reflect both differences in instrument hardware as well as normalization techniques. These differences will need to be considered when developing universal segmentation algorithms and normalization strategies between devices.
Keywords: 688 retina •
549 image processing •
551 imaging/image analysis: non-clinical