To evaluate macular thickness measurement reproducibility using 6 new spectral-domain (SD) optical coherence tomography (OCT) devices and 1 time-domain OCT device
Prospective, observational study.
setting: Clinical practice. study population: Macular thickness was assessed in 18 randomly chosen consecutive eyes of 18 healthy volunteers by 2 masked operators using 6 SD OCT devices and 1 time-domain OCT device. main outcome measures: Intraoperator and interoperator reproducibility of OCT-measured central macular thickness.
Mean macular thickness ranged from 172.77 to 272.87 μm. Intraobserver Intraclass coefficient correlation ranged from 0.75 to 0.96. Intraobserver coefficient of variation ranged from 0.44 to 2.75. In Bland-Altman analysis, interoperator mean difference ranged from 0.22 to −9.69. An analysis of variance used for repeated measurement was statistically significant for instruments ( P < .001) and operators ( P = .04), but not for instruments × operators ( P = .32).
The 7 OCT devices presented differing macular thickness measurement values; the lowest value was with Copernicus, and the highest was with Spectralis HRA+OCT (Heidelberg Engineering). The new generation of SD OCT devices has good intraoperator reproducibility, but the Spectralis presents the highest reproducibility together with the best interoperator agreement. The software, and in particular the algorithm, used is the most important factor regarding reproducibility differences in macular thickness measurement. The main result of our study is that macular thickness absolute value differs for each device. For this reason, the devices are not interchangeable.
Optical Coherence Tomography (OCT) is a valuable technique for the detection and monitoring of a variety of macular diseases. A new generation of OCT devices recently was introduced and currently is in clinical use. These devices, called spectral-domain (SD) or Fourier-domain OCT, use a spectrometer that can sample more than 20 000 A-scans per second and, therefore, collect far more data than is possible using the time-domain system, which functions at approximately 500 A-scans per second.
SD OCT is able to gather in-depth data from the spectra of the OCT signal. The SD OCT technique offers better axial resolution (5 μm) and increases the speed of data collection by a factor of more than 40 times. The increased speed of SD OCT means there is less eye movement during scans, thus resulting in more stable images. The image stack across the macula can be processed to produce 3-dimensional structural representations. Segmentation of 3-dimensional images permits better visualization of the retinal layers than visualization with time-domain OCT and also allows 3-dimensional eye mapping.
Measurement reproducibility is an essential parameter when determining the clinical usefulness of a device. Previous studies have investigated the reproducibility of measurements obtained using time-domain OCT. These studies showed that OCT can provide highly reproducible measurements of retinal layer thicknesses.
Spectral OCT/SLO (Opko/OTI; Ophthalmic Technologies, Inc, Toronto, Canada), 3D OCT-1000 (Topcon, GB Ltd, Newbury, Berkshire, United Kingdom), RTVue-100 (Optovue, Fremont, California, USA), Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, California, USA), SOCT Copernicus (Reichert, Inc, Depew, New York, USA), the Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Germany) are just some of the new devices on the market that have been used in our ophthalmology unit. A recent study evaluated the intersession reproducibility of 5 of the 6 commercially available SD OCT instruments and 1 time-domain OCT instrument. Our goal was to determine the intraoperator and interoperator reproducibility of macular thickness measurement in normal eyes using 6 commercially available SD OCT devices as well as a standard time-domain OCT instrument.
Macular thickness was assessed in 18 randomly chosen consecutive eyes of 18 healthy volunteers from the staff of our department by 2 masked operators, L.P. (A) and E.M. (B), with similar practical OCT experience. Six SD OCT devices were used: Spectral OCT/SLO (Opko/OTI; software version November 2007); 3D OCT-1000 (version 2.12); RTVue-100, (software version 3.0); Cirrus HD-OCT (software version 3.0); SOCT Copernicus (software version 3.03); Heidelberg Spectralis HRA+OCT (software version 3.1.5); and 1 time-domain OCT device was used: Stratus OCT (software version 4.0.1; Carl Zeiss Meditec, Inc).
Only subjects with no history or evidence of intraocular surgery, retinal disease, or glaucoma, and with refractive error less than 2 diopters qualified as normal. Both observers repeated 2 consecutive measurements on 1 eye of each subject during the same day at the same time for each instrument used. All subjects underwent OCT imaging using each of the 7 devices at various times from 2007 through 2008.
Central foveal thickness was determined automatically and was analyzed by OCT software. The pupil was not dilated. In all OCT maps, automated macular thickness detection was performed without manual operator adjustment. Stratus OCT used the fast macular thickness protocol, consisting of 6 consecutive 6-mm radial line scans centered on the macula in 1.5 seconds, each containing 128 A-scans. Spectral OCT/SLO Opko/OTI uses 64 raster lines consisting of 512 A-scans every 2 seconds in a 5 × 5-mm area, 27 000 A-scans per second; 3D OCT-1000 Topcon uses 128 raster lines consisting of 512 A-scans per second in a 6 × 6-mm area (512 × 128) in 3.6 seconds, 18 000 A-scans per second; RTVue-100 Optovue uses a dense grid of 34 B-scans, consisting of 558 A-scans in a 5 × 5-mm area of the central macula to collect a number of oversampled points in 0.98 seconds, 26 000 A-scans per second; Cirrus HD OCT Zeiss uses 128 raster lines consisting of 512 A-scans per second, in a 6 × 6 mm area (128 × 512) in approximately 2.5 seconds, 27 000 A-scans per second; SOCT Copernicus Reichert uses 50 raster lines consisting of 743 A-scans per 1.5 seconds in a 7 × 7-mm area (50 × 743), 25 000 A-scans per second; Spectralis HRA+OCT Heidelberg uses 73 raster lines consisting of 40 000 A-scans per second in a 5.8 × 4.35-mm area in 5 seconds, factor for scan averaging: 6. Intraobserver reproducibility was evaluated by randomly using intraclass coefficient correlation (ICC) between instruments and by a coefficient of variation (CV) between the 2 measurements carried out by each operator. Interobserver variability was evaluated by Bland-Altman analysis per instrument, using the mean values of the repeated measurements of the 2 operators and the corrected standard deviation between the 2 repeated measurements per operator. To assess instrument-to-instrument and operator-to-operator variability, we used an analysis of variance to determine instrument-to-instrument reproducibility for repeated measurements with 2 within factors (instruments–operator); a post hoc test for differences among instruments also was evaluated.
Mean age was 37 years (range, 29 to 56 years); there were 6 men and 12 women. Data were normally distributed. In all OCT maps, automated macular thickness detection was performed correctly without manual operator adjustment.
Mean macular thickness for observers A and B are reported in Table 1 . The ICC (95% confidence interval) for observer A ranged from 0.75 (RTVue-100) to 0.96 (Spectralis HRA+OCT and Cirrus HD OCT). The ICC for observer B was slightly higher, but showed the same trend. The lowest ICC was for RTVue-100 and the highest was for Spectralis HRA+OCT ( Table 2 ). The CV for operator A ranged from 2.75 (Stratus OCT) to 0.44 (Spectralis HRA+OCT); the CV for operator B ranged from 2.66 (Stratus OCT) to 0.43 (Spectralis HRA+OCT; Table 2 ).
|Instrument||Operator 1||Operator 2|
|Mean a||SD||Mean a||SD|
|Stratus OCT (1)||202.88||13.56||206.27||18|
|Spectral OCT/SLO (2)||213.02||10.03||215.02||10.39|
|3D OCT-1000 (3)||224.41||18.19||225.30||25.66|
|Cirrus HD OCT (5)||253.94||9.73||253.72||9.75|
|SOCT Copernicus (5)||172.66||7.92||172.88||8.51|
|Spectralis HRA+OCT (5)||273.19||8.29||272.55||8.88|
|Instrument||Interoperator Reproducibility (ICC)||Intraoperator Reproducibility|
|Operator A (ICC)||Operator B (ICC)||Operator A (CV)||Operator B (CV)|
|Stratus OCT||0.86 (67–94)||0.88 (72–95)||0.92 (80–96)||2.75||2.66|
|Spectral OCT/SLO||0.81 (56–92)||0.87 (70–95)||0.90 (75–96)||1.83||1.6|
|3D OCT-1000||0.78 (46–92)||0.77 (49–90)||0.84 (63–93)||1.96||2.08|
|RTVue-100||0.82 (59–93)||0.75 (43–90)||0.76 (47–90)||1.11||1.04|
|Cirrus HD OCT||0.89 (74–96)||0.96 (89–98)||0.93 (83–97)||0.61||0.64|
|SOCT Copernicus||0.84 (63–94)||0.76 (47–90)||0.81 (57–92)||2.11||1.81|
|Spectralis HRA+OCT||0.96 (90–99)||0.96 (91–98)||0.97 (93–99)||0.44||0.43|