The rapid acquisition of B-scans possible with spectral-domain optical coherence tomography (SD-OCT) allows dense scanning of the retina, with the resulting ability to create 3-D representations of retinal morphology. One of the advantages of high-speed scanning is that it improves the diagnostician's ability to detect clinically relevant features of retinal disease.1 Dense scanning also allows creation of more accurate retinal thickness maps because the technology is less dependent on interpolation of data between scans.

Despite the rapidity of SD-OCT scanning, however, patient eye movement can still occur during acquisition of the dense datasets used to construct 3-D images. Even in systems that track eye movement, patients with poor fixation may cause the system to “time out,” limiting the scanning density and the amount of averaging that is possible.

In the clinical trial setting, additional challenges are presented by errors in retinal boundary detection with SD-OCT, which still occur despite improvements in segmentation software. Manual correction of the dense datasets generated by SD-OCT can be time-consuming and impractical, even at reading centers.

To better define the optimal scan density for generating reliable retinal thickness maps, we investigated the relationship between B-scan density and retinal thickness measurements obtained by SD-OCT in eyes with retinal disease.2

FULL AND SUBSET SCANS
Investigators at Doheny Retina Institute collected data from 115 eyes of 115 patients who underwent volume OCT imaging with Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA) using the 512 x 128 horizontal raster protocol. Consecutive eyes rated as having accurately segmented boundaries were included. The patient population represented a fairly wide distribution of retinal diseases (Table).

Raw imaging data, including the location of automated retinal boundaries, were transferred from the HDOCT instrument to validated OCT viewing and grading software developed at the Doheny Image Reading Center.

Retinal thickness maps similar to those produced by the Cirrus HD-OCT were generated for each eye using all 128 B-scans and using less dense subsets of scans — every other scan, every fourth scan, and so on, up to every 16th scan (Figure 1).

Retinal thickness measurements derived using these subsets of scans were compared to measurements using the full 128 B-scans. We computed differences between the full and subset scans for various subfields, including the foveal central subfield (FCS), an important measurement in clinical trials, as well as total macular volume.

RESULTS
The mean absolute error in retinal thickness measurements increased as the density and number of B-scans was decreased (Figures 2 and 3).,/p>

Using every fourth scan (ie, 32 scans spaced 188 μm apart), the mean thickness at the FCS differed by less than 1 μm from the full 128 scans, and the maximum error was only 3 μm. This equated to a mean percentage error of less than 0.5% and a maximum error of less than 1%. When the density was decreased to every 16th scan (eight scans 750 µm apart), the mean error at the FCS increased to 7 µm (2.5%) and the maximum error to 71 μm (15.5%).

Total macular volume was more resistant to change as B-scan density was reduced. Even at the lowest scan density of every 16th scan, the mean absolute error was 0.1 mm3 (0.18%) and maximum error was 0.13 mm3 (1.7%).

Disease diagnosis was not correlated with any observed measurement differences. However, there were relatively small numbers of patients in each disease category (Table), and the study was not powered to evaluate this issue.

Reduced B-scan density appeared to overestimate retinal thickness at lower levels of retinal thickness and underestimate it at higher levels.

DISCUSSION AND CONCLUSIONS
This study was limited in certain aspects. We evaluated only retinal thickness, and it is possible that different, and possibly higher, scanning densities will be required for study of other sub-compartments, for example in trying to quantify subretinal fluid or drusen.

In addition, higher density scanning may still be required to avoid reducing sensitivity for detecting retinal features of interest. The design of clinical trials may require different densities of acquisition for the purpose of qualitative vs. quantitative assessments.

In summary, we observed that B-scan density in volume SD-OCT acquisitions can be reduced to 32 horizontal B-scans over a 6-mm zone (ie, spaced 188 µm apart) with minimal change in calculated retinal thickness measurements. This information may be of value in the design of scanning protocols for SD-OCT use in future clinical trials and ultimately in clinical practice.

Srinivas R. Sadda, MD, is an Associate Professor of Ophthalmology at the Doheny Eye Institute and University of Southern California in Los Angeles. He shares in royalties from intellectual property licensed to Topcon Medical Systems from Doheny Eye Institute and has served as a consultant for Heidelberg Engineering. The Doheny Image Reading Center has received research support from Carl Zeiss Meditec. He can be contacted at +1 323 442 6522; or via e-mail at: SSadda@doheny.org.

  1. Keane PA, Bhatti RA, Brubaker JW, Liakopoulos S, Sadda SR, Walsh AC. Comparison of clinically relevant findings from high-speed Fourier-domain and conventional time-domain optical coherence tomography. Am J Ophthalmol. 2009;148(2):242– 248.
  2. Sadda S, Keane P, Ouyang Y, Updike J, Walsh A. Impact of scanning density on measurements from spectral domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2009 Sep 24. [Epub ahead of print]