Abstract
purpose. To compare central retinal thickness (CRT) measurements in healthy eyes by different commercially available OCT instruments and to compare the intersession reproducibility of such measurements.
methods. Six different OCT instruments (Stratus OCT [Carl Zeiss Meditec, Inc. Dublin, CA], SOCT Copernicus [Reichert/Optopol Technology, Inc., Depew, NY], Spectral OCT/SLO [Opko/OTI, Inc., Miami, FL], RTVue-100 [Optovue Corp., Fremont, CA], Spectralis HRA+OCT [Heidelberg Engineering, Inc., Heidelberg, Germany], and Cirrus HD-OCT [Carl Zeiss Meditec, Inc.]) were used to assess CRT in both eyes of healthy subjects. Measurements were performed in two different sessions on the same day with each of the systems. From these measurements, the mean CRT was calculated. For the assessment of the intersession reproducibility of the instruments, we calculated the coefficient of the variation of test–retest variation.
results. Twenty healthy subjects were included in the study. Compared with the Stratus OCT all spectral OCT instruments showed significantly higher CRTs. The Spectralis HRA+OCT and Cirrus HD-OCT showed similar CRT values but significantly higher values than did all other instruments. The coefficients of variation for repeated measurements was 3.33% for the Stratus OCT, 0.46% for the Spectralis HRA+OCT, 3.09% for the Cirrus HD-OCT, 2.23% for the OCT/SLO, 2.77% for the RTVue-100 OCT, and for the SOCT 3.5%, respectively.
discussion. The six OCT systems provided different results for CRT. The measurements with the Stratus OCT showed the lowest thicknesses, whereas those with the Cirrus HD-OCT and Spectralis HRA+OCT yielded the highest ones. These discrepancies can be explained by the differences in the retinal segmentation algorithms used by the various OCT systems. Whereas the Spectralis HRA+OCT and Cirrus HD-OCT include the RPE layer in the retinal segmentation, the other instruments do not. The data imply that the different OCT systems cannot be used interchangeably for the measurement of macular thickness.
One of the most exciting developments in ophthalmic imaging is probably optical coherence tomography (OCT), which was introduced in 1991.
1 2 3 4 Being a part of clinical practice since 1995, OCT may evolve as a noninvasive investigation that replaces or is complementary to fluorescein angiography. Future therapeutic options are in discussion.
2 3 5 6 7
Today, the conventional time-domain OCT image of the posterior pole centered on fixation is well understood. A hyperreflective (color coded red) line in the midthickness of the trace is a central reference point and represents the retinal pigment epithelium (RPE)-choriocapillaris complex. Posterior to this structure, the signal weakens due to attenuation. Anterior to the RPE/choriocapillaris, the retina is seen as a layer of lower reflectivity (color coded green-yellow). Foveal depression and the optic disc are clearly visible in physiological conditions.
Descriptions of the different intraretinal layers in OCT images have been published in detail. Thoth et al.
8 described the nerve fiber layer as a hyperreflective layer, usually visible where the layer is thickest: nasal to the fovea in horizontal sections and superior and inferior to the fovea in vertical sections.
8 Others (e.g., Chauhan and Marshall
9 ) have questioned this description, and it is probably only acceptable for describing retinal thickness as the distance between the hyperreflective layer representing the RPE/photoreceptor complex and the vitreoretinal interface.
9 In these cases, normal foveal thickness measured by time-domain OCT is approximately 210 μm.
10 11 The posterior hyaloid interface may be visible as a thin, hyperreflective (color coded red-white) structure against the dark hyporeflective vitreous cavity. However, this layer is only discernible if there is vitreoretinal separation,
8 and it has to be mentioned that there is still controversy regarding the correlation of the outer OCT hyperreflective bands with anatomic findings.
In recent years, the conventional time-domain OCT has played an important role as a diagnostic tool in monitoring patients with macular disorders. It is used in various studies to assess morphologic changes during therapy by analyzing macular thickness.
12 13 14 When observing patients over time, it is important to assess changes in macular thickness correctly and to be able to compare baseline findings with follow-up measurements. In this context, the widely used OCT-3 Stratus model, released in 2002, and its updated software version 4.0 have well-known limits. The conventional time-domain OCT often produces false macular thickness maps in patients with macular disorders, because the retinal thickness algorithm is not always able to identify the inner border of the RPE and/or the inner limiting membrane (ILM). One reason for this deficiency may be the severe morphologic changes in the retina and RPE of these patients.
Current advances (e.g., the introduction of Fourier analysis; spectral OCT) have made high resolution and fast scanning speed possible,
15 16 17 with resolution being up to five times higher and imaging speed up to 60 times faster than in conventional time-domain OCT.
16 High-resolution OCT makes differentiation of as many as 11 structural characteristics within the retina
15 18 possible.
The purpose of our study was to report macular thickness measurements in healthy eyes by using different commercially available OCT instruments and to evaluate the reproducibility of test results with these instruments. In addition, we analyzed differences in mean macular thickness measurements for each of the different OCT instruments. We believe that this should be the first step in analyzing macular thickness measurements. The next most important step is analyzing patients with different retinal diseases to assess the limits and errors of OCT thickness measurements.
Subjects with clear media in both eyes and normal retinal status were recruited for our study. Inclusion criteria were an age of 18 years or older, best corrected visual acuity (BCVA) of ≥20/20, refractive error of ≤3 D, and no history or evidence of either systemic or ophthalmic disease. In eligible subjects, both eyes were studied.
Before OCT examination, all subjects underwent a complete ophthalmic examination including a detailed medical and ocular history, BCVA with ETDRS charts, binocular ophthalmoscopy, and color fundus photography (FF 450 plus; Carl Zeiss Meditec, Jena, Germany). Six different OCT instruments
(Table 1)were used to assess central retinal thickness (CRT, mean thickness in the central 1000-μm diameter area). Each subject underwent two acquisition sessions on each of the six OCT instruments within 2 hours on the same day. In the first series, each subject was taken through one acquisition session for both eyes (first the right eye then the left) with each instrument used in random order. Thereafter, a second series of acquisition sessions was undertaken in the same order. The OCT instruments were operated by the same trained operator. Only OCT scans that were of sufficient quality (signal ≥ 50% of maximum strength, absence of imaging artifacts, or distortions) were used. Replicates were only taken if the OCT scans were of insufficient quality. We attempted to use similar acquisition protocols on each instrument
(Table 2) , but especially for the OCT-3 Stratus, no volume scan protocol is available. Therefore, we used the Fast Macular Thickness map protocol. For acquisition of the volume scans with the Spectralis HRA+OCT the unique feature of real-time averaging of line scans was used. Six scans were averaged to produce one line scan to be used as the volume scan.
Before examination, the pupil of the study eye was dilated to at least 6 mm diameter with drops containing 0.5% tropicamide and 2.5% phenylephrine. The procedure was similar for each of the instruments. After the acquisition protocol and scan procedure were explained to them in detail, the subjects were placed in front of the OCT instrument and asked to fixate on an internal fixation target. After the focus was adjusted and a good central fixation obtained, the scanning process was initiated.
From the OCT examinations, thickness maps were calculated with the built-in analysis software of the OCT. For analysis, the mean and standard deviations was calculated for the CRT measured with each of the six different instruments, separately for the right and left eyes. Differences between the eyes were assessed by the paired Student’s
t-test. To assess the differences between the instruments, we used only measurements of the right eye of each subject. The paired Student’s
t-test was used to compare the CRT measurements from all instruments. All
P-values were adjusted for multiple testing according to Holm.
19 For the assessment of intersession repeatability, we recalculated the coefficient of variation (CV) from
equation 1to determine test–retest variation.
20 \[cv\ {=}\ \frac{\sqrt{\frac{{{\sum}}\ (x_{i}^{1}\ {-}\ x_{i}^{2})^{2}}{2\ {\cdot}\ n}}}{\frac{{{\sum}}\ x_{i}^{1}}{n}}\]
where
x i 1 is the result from the first series,
x i 2 is the result from the second series, and
n is the number of eyes.
This study was performed with the informed consent of the participants. It was conducted under a protocol approved by the local institutional review board, in accordance with the ethics stated in the Declaration of Helsinki (1964) and with the recommendations of the local ethics committee.
Forty eyes of 20 healthy subjects (11 women, 9 men) aged from 25.0 to 63.0 years (37.1 ± 12.8) were included in the study. The refractive error ranged between −2.0 and +1.0 D (mean, 0.34 ± 0.82 D). Measurement of CRT was possible in all 40 eyes. All the OCT scans showed signal strength > 70% of maximum strength, and no errors in the layer segmentation were observed.
Table 3shows the mean CRT separated for right and left eyes for all six instruments. No significant differences in CRT were found between the right and left eyes.
Table 4presents the comparison of the measurements obtained with the six instruments (significances above the diagonal and differences below). The differences in CRT ranged between 2 and 77 μm. Compared with the time-domain instrument Stratus OCT, all spectral OCT instruments showed significantly higher CRTs. The Spectralis HRA+OCT and Cirrus HD-OCT showed similar CRTs but significantly higher ones than all the other instruments. In addition, for comparison of CRT measurements on the different OCT machines, Bland Altman plots were calculated
(Fig. 1) .
To quantify the intersession reproducibility of the instruments, we performed repeated measurements.
Table 4shows the CRTs for the first and second measurements, as well as the intersession repeatability (CV) for all instruments. The intersession repeatability (CV) was 3.33% for the Stratus OCT, 0.46% for the Spectralis HRA+OCT, 3.09% for the Cirrus HD-OCT, 2.23% for the OCT/SLO, 2.77% for the RTVue-100 OCT, and 3.5% for the SOCT. Compared with the Stratus OCT, only the Spectralis HRA+OCT showed a significantly lower variation between the two measurements. All other instruments had similar reproducibility
(Table 5) .
In this study, we compared the measurements of CRT in healthy subjects by six different commercially available OCT instruments. These included one time-domain OCT and five spectral OCT systems. In our study, the measurements of CRT in a group of healthy subjects ranged between 212 ± 19 and 289 ± 16 μm. Because only healthy subjects with good fixation were included, the OCT scans were of excellent quality and the segmentation of retinal layers showed no errors. In patients with retinal diseases, this may be different and could include additional variability. The time-domain results differ significantly between the different OCT instruments. The CRT measurements can be subdivided into three different range groups. The Stratus OCT produced the lowest CRT; the SOCT Copernicus, Spectral OCT/SLO, and the RTVue-100 produced midrange values; and the Spectralis HRA+OCT and Cirrus HD-OCT produced the highest ones. The differences between the instruments include different methods of scan acquisition, of segmentation of the retinal borders, and of sampling the measurement points and probably different estimates for the optical indices of the retina. In addition, there are important differences between the algorithms for alignment and registration of the OCT scans.
The acquisition protocols differ substantially between the instruments
(Table 2) . The Stratus OCT system uses only six radial lines with a total of 768 A-scans to produce a thickness map with a diameter of 6 mm. Because the density of the measuring points is dependent on the distance from the center, only measurements inside the central 1000-μm diameter area are based on a sufficient number of A-scans (128 A scans). The other instruments use rectangular scan patterns resulting in a uniform density of A-scans within the scan area. However, the number of A-scans per square millimeter differs considerably among the instruments. The density varies between 1428 A-scans/mm
2 (Cirrus HD-OCT) and 524 A-scans/mm
2 (Spectral OCT/SLO). The differences in the number of A-scans per square millimeter may have an influence on the CRT, but this influence can only be evaluated by repeated measurements with the same instrument with different scan densities in a future study. The acquisition time of the volume scans ranges between 1.5 and 5.0 seconds. The longest acquisition time was observed with the Spectralis HRA+OCT. This long acquisition time is due to the real-time averaging feature of the system. But the averaging feature is combined with a real-time tracking system for eye movements, and so the long acquisition time had no negative influence on repeatability.
The segmentation software of all the instruments identifies different hyperreflective structures in each line scan. The segmentation of the inner retinal border is not different among the instruments. All instruments identify the vitreoretinal interface as the inner retinal border. The segmentation of the outer retinal border differs among the instruments significantly. The Stratus OCT system image the outer retinal layers (RPE-photoreceptor complex) as two hyperreflective bands. The segmentation software of the Stratus OCT system uses the inner hyperreflective band for segmentation. The new spectral OCT systems image the outer retinal layers typically as three hyperreflective bands. The most inner of these hyperreflective bands has the lowest reflectivity and cannot be imaged with the Stratus OCT system. The bands may correspond to the external limiting membrane, the junction of the photoreceptor outer segments (OS) and inner segments (IS), and the RPE. The SOCT Copernicus, Spectral OCT/SLO, and RTVue-100 use the second inner hyperreflective band as the outer border of the retina. The Cirrus HD-OCT and the Spectralis HRA+OCT identify the most outer reflective band as the outer border of the retina
(Fig. 2) . Thus, the Stratus OCT system would be expected to yield CRTs that are lower than the ones from all other instruments, whereas Cirrus HD-OCT and Spectralis HRA+OCT system should yield the highest measurements. Some of the instruments allow manual correction of the automated segmentation. In this study, we did not use the manual adjustment to change the segmentation layers. It may be possible to obtain similar results by adjusting the segmentation, but this was not the subject of our study.
In this study, we also measured and compared the repeatability indices of the Stratus OCT, SOCT Copernicus, Spectral OCT/SLO, and RTVue-100, Spectralis HRA+OCT, and Cirrus HD-OCT systems within the same population of healthy subjects. For OCT scans obtained by the same experienced operator with adequate signal strength and accurate segmentation, all systems produced measures that reflected low variance. The CV of repeated measurements ranged between 0.46% with the Spectralis HRA+OCT and 3.5% with the SOCT Copernicus system. The high repeatability of the Spectralis HRA+OCT measurements is most probably related to the unique feature of the system that allows eye tracking during the scanning process as well as automatic recognition of the exact same scan location for follow-up examination. By using this feature for all follow-up scans with the Spectralis HRA+OCT, we could minimize extrinsic factors, such as patient fixation and the operator’s ability to consistently place the macular grid over the same points during each scan.
Currently, several studies on the reproducibility of time-domain OCT systems but only scant data on reproducibility of spectral domain OCT systems have been published. In 2001, Massin et al.
7 reported results using the first commercially available time-domain OCT. In their study, retinal thickness measurements in nine ETDRS areas of 10 healthy eyes were analyzed. Interclass correlation coefficients ranged from 0.89 to 0.99. In addition, they tested reproducibility in patients with clinically significant macular edema. In these diabetic patients, the interclass correlation coefficients were always larger than 0.98 and the reproducibility was ±6%. Muscat et al.
21 and Koozekanani et al.
22 also demonstrated good reproducibility, with overall coefficients between 1% and 2% and an expected variation of less than 11 μm between measurements. Gürses-Ozden et al.
23 found acceptable reproducibility in healthy subjects. The Stratus OCT measurement yielded a CV of 5.8% using the Fast Macular Thickness scan and a CV of 4.7% using the Radial Line Scan. In 2006, Polito et al.
24 reported CVs from 1.68% to 6.63% in a healthy group and from 4.84% to 8.33% in a diabetic group. A recent study compared macular thickness measurements and their repeatability in the Stratus OCT and the Cirrus HD-OCT system in patients with diabetic macular edema (DME).
25 In addition, a study of healthy subjects comparing the Stratus OCT and the 3D OCT (Topcon, Tokyo, Japan) was performed by Leung et al.
26 We found that all these data on CRT measurement and repeatability were comparable with ours.
Beside the CRT measurements, the high resolution of every spectral OCT scan gives important information about the structure of the retina and allows analysis of more layers than do the Stratus OCT scans. In patients with macula disorders, high-resolution retinal thickness maps may provide additional information about the stage of the disease and allow more precise comparison of follow-up OCT results with baseline findings.
27
In summary, we found that the OCT systems provided different values for CRT: Measurements with the Stratus OCT showed the lowest values, whereas measurements with the Cirrus HD-OCT and Spectralis HRA+OCT yielded the highest ones. These discrepancies were most probably based on differences in retinal segmentation algorithms used by the various OCT systems. These data imply that the different OCT systems cannot be used interchangeably for the measurement of macular thickness. The unique feature of the Spectralis HRA+OCT system, which allows automatic recognition of the exact same scan location, resulted in the best intersession reproducibility of measurements.