Article Text
PDF
Abstract
Aim Spectral domain optical coherence tomography (SD OCT) is of increasing importance and is gradually replacing time domain OCT (TD OCT). Our aim was to determine a formula to convert Stratus OCT (TD OCT) to Cirrus OCT (SD OCT) retinal thickness.
Methods Central retinal thickness (CRT) and retinal volume (RV) were obtained by the macular thickness program of Stratus OCT and the cube 512×128 program of Cirrus OCT in patients with exudative age-related macular degeneration (AMD). Algorithm line failures were corrected. A linear model with Stratus OCT CRT as fixed factor and Cirrus OCT CRT as dependent variable was applied to calculate the conversion formula.
Results OCT examinations of 104 eyes of 104 patients were reviewed and corrected when necessary. Stratus and Cirrus OCT CRT were significantly correlated (p<0.0001). For CRT the formula Cirrus CRT=58.63+0.94× Stratus CRT was calculated. The correlation was significantly influenced by the height of the CRT values (p<0.0001), but not by whether correction was necessary. For RV the formula Cirrus OCT RV=3.098+0.98× Stratus OCT RV was calculated.
Conclusion Stratus OCT and Cirrus OCT use a different posterior reference line within the hyper-reflective band of the outer retina. Therefore a conversion formula is necessary to compare Stratus and Cirrus OCT CRT values, and this has been determined in our study.
- Age-related macular degeneration
- retinal thickness
- optical coherence tomography
- spectral domain OCT
- choroidal neovascularisation, retina
- neovascularisation
Statistics from Altmetric.com
- Age-related macular degeneration
- retinal thickness
- optical coherence tomography
- spectral domain OCT
- choroidal neovascularisation, retina
- neovascularisation
Introduction
Optical coherence tomography (OCT) has become one of the most important examinations in the diagnosis and follow-up of exudative age-related macular degeneration (AMD). New treatment and re-treatment regimens with inhibitors of the vascular endothelial growth factor (anti-VEGF therapy) in particular are based on differences in retinal thickness.1 Furthermore, OCT RT measurements provide data representing the activity of a lesion,2 3 which might also be used to demonstrate the effect of a treatment in clinical trials.
Until recently Stratus OCT representing time domain technology was the only commercially available OCT machine. A modern OCT technology, the spectral domain technology, has now become available. This new technology is offered by several companies and provides a series of advantages, such as higher resolution, much faster acquisition speed and a different acquisition pattern. Whereas in Stratus OCT only six radial lines are recorded and the retinal thickness of the spaces in-between the lines is interpolated, in spectral domain OCT a raster of a series of horizontal lines is recorded, offering directly measured data on retinal thickness. Because of these advantages, spectral domain OCT machines are becoming more widely used. However, it would certainly be of advantage for retinal thickness measurements performed by Stratus OCT to be further available, especially in studies with a longer follow up. Conversion formulae are required because a different posterior reference line is used by Stratus OCT and because of the use of different spectral domain OCT machines.
In this study we calculated formulae to convert central retinal thickness (CRT) and retinal volume (RV) data, the most widespread used parameters offered by the OCT software, between Stratus OCT and Cirrus OCT, a representative of spectral domain technology.
Methods
The median age of the 104 patients included was 78 (range 52–95) years, 65.4% were female, 39.8% were treatment naïve and 70.2% were receiving anti-VEGF treatment. The OCT scans of 104 eyes of 104 patients suffering from neovascular AMD were reviewed. The scans were available from a former study dealing with differences of threshold algorithm line quality between Stratus and Cirrus OCT.4 The CRT (mean value of retinal thickness in the central area of 1 mm diameter) and the retinal volume (RV) within a circle of 6 mm diameter were recorded.
The standard programs for retinal thickness measurements with the OCT were performed: (1) the macular thickness program consisting of six radial lines through the centre of the foveal region of Stratus OCT (model 3000, software version 5.0; Carl Zeiss Meditec Inc., Dublin, California, USA); (2) the cube 512×128 program consisting of 128 horizontal lines of 512 A-scans of the Cirrus OCT (software version 3.0; Carl Zeiss Meditec Inc.). Threshold algorithm line errors (identified by two independent observers, IK and SH, and graded according to the severity of the failures in the previous study) were corrected by an experienced examiner, IK, with the help of the built-in software of both OCT machines (software version 5.0 of Stratus OCT and 3.0 of Cirrus OCT). Erroneous lines were redrawn on the screen with the help of the computer mouse. Eyes with segmentation failures that could not be corrected were excluded from analyses.
Statistical analyses
The conversion formulae between Cirrus CRT and Stratus CRT, as well as between Cirrus RV and Stratus RV, were calculated by a linear model with the measurements of Cirrus as dependent variable and the measurements of Stratus as fixed factor. To identify an eventual difference between corrected (at least correction for Stratus or Cirrus) and uncorrected measurements, we calculated a similar model but with additional correction yes/no and the interaction term between the independent variables as a fixed factor.
The aim of these analyses was to illustrate how (and with what error) Cirrus and Stratus measurements may be converted. The residual SE was calculated as a measure of how precise the values of one device (Cirrus or Stratus) can be expressed in terms of the other device (Stratus or Cirrus).
Analyses were done with the statistical program R 2.8.0, R Foundation for Statistical Computing, Vienna, Austria. For all analyses, the significance level was set to 0.05.
Results
The scans of 72 eyes in Stratus OCT and 31 eyes in Cirrus OCT that exhibited algorithm line errors were corrected by IK. Two patients showed algorithm line errors that could not be corrected because the retina surface could not be identified. The mean values of CRT±SD of the remaining 112 eyes, based either on correctly automatically set or manually corrected lines, were 282.6±82.4 (143–558) μm for Stratus OCT and 322.0±90.6 (164–642) μm for Cirrus OCT. RV values of 7.3±1.1 (4.6–10.44) mm3 and 10.3±1.2 (6.6–14.8) mm3 were calculated for Stratus and Cirrus OCT, respectively. The mean difference between Cirrus and Stratus OCT was 41.1±46.6 (–110–169) μm and 3.0±0.4 (2–4.8) mm3 for CRT and RV, respectively.
The linear model revealed a significantly positive correlation between Cirrus CRT and Stratus CRT (p<0.0001, formula (95% CI): Cirrus CRT=58.63 (25.84 to 91.41)+0.94 (0.83 to 1.05)× Stratus CRT, degrees of freedom=100). The linear model with Cirrus CRT as dependent variable and Stratus CRT, correction yes/no and the interaction between Stratus CRT and correction revealed a significant effect of Stratus CRT (p<0.0001), but not of correction (p=0.6). The interaction term was not significant either (p=0.4). An example showing the differences of positioning of the posterior line between Stratus and Cirrus OCT is presented in figure 1. In this case the Stratus CRT was 269 μm, the calculated Cirrus CRT was 311.4 μm and the definite Cirrus CRT was 311.0 μm.
Central horizontal scans of Stratus optical coherence tomography (OCT) (top) and Cirrus OCT (bottom) of a female patient (age 63 years); occult type 2 lesions are present. The posterior reference line is marked by arrow (on the surface of the hyper-reflective band in Stratus OCT and at the third hyper-reflective line in Cirrus OCT. The different appearance of the scans is caused by alignment in Stratus OCT.
The correlation between Cirrus RV and Stratus RV was also significantly positive (p<0.0001, formula (95% CI): Cirrus RV=3.098 (2.497 to 3.700)+0.982 (0.902 to 1.063)× Stratus RV, degrees of freedom=101). The linear model with Cirrus RV as dependent variable and Stratus RV, correction yes/no and the interaction between Stratus RV and correction revealed a significant effect of Stratus RV (p<0.0001), but not of correction (p=0.7). The interaction term was not significant either (p=0.8). The correlations between Stratus and Cirrus OCT are presented in figure 2 (CRT) and figure 3 (RV).
Correlation between Cirrus central retinal thickness (CRT) and Stratus CRT. Triangle, correction of Stratus; inverted triangle, correction of Cirrus; square, correction of Stratus+Cirrus; circle, no correction. Black line, regression line Cirrus CRT versus Stratus CRT. The inner 95% CI (dotted lines) represent the prediction of the mean values; the outer 95% CI (dashed lines) represent the prediction of the individual values. Grey line, regression line of the uncorrected measurements. The distribution of these data points shows that the correlation between Stratus and Cirrus CRT was independent of whether correction was necessary or not.
Correlation between Cirrus retinal volume (RV) and Stratus RV. Triangle, correction of Stratus; inverted triangle, correction of Cirrus; square, correction of Stratus+Cirrus; circle, no correction. Black line, regression line Cirrus RV versus Stratus RV. The inner 95% CI (dotted lines) represent the prediction of the mean values; the outer 95% CI (dashed lines) represent the prediction of the individual values. Grey line, regression line of the uncorrected measurements. The data points are crowded along the regression line: the distribution revealed no influence of whether the values had to be corrected or not.
Discussion
Retinal thickness is defined as the distance between the two most commonly automatically set threshold algorithm lines at the retinal surface and the hyper-reflectivity of the outer retina corresponding to the pigment epithelium choriocapillaris complex. The values of Stratus OCT and Cirrus OCT of CRT and RV are different because both machines position the posterior boundary line at different locations within the posterior hyper-reflective layer (Stratus OCT at the surface of the most anterior line visible in Stratus OCT and corresponding to the area of the junction of the inner and outer segments of the photoreceptors; Cirrus OCT using the third and most posterior line as reference line).5 After threshold algorithm line error correction, data from 102 eyes were available and conversion formulae could be calculated, composed of a constant added to a multiple of the value of the examination of the other OCT machine. These formulae take into account that the differences between CRT obtained by Stratus and Cirrus OCT are dependent on the height of retinal thickness. The statistical analyses confirmed the significant influence of the height of the CRT values, whereas whether correction was necessary or not did not reach significance. Furthermore, neovascular AMD is located at the area of the retinal pigment epithelium and therefore the distance between the hyper-reflective lines might be influenced by the pathology itself. Therefore, a conversion formula calculated from values obtained in health eyes might be not applicable in eyes with neovascular AMD. Carpineto and co-workers6 found in 40 healthy subjects a linear relation between Spectralis OCT (a spectral domain technology OCT machine similar to Cirrus OCT) and Stratus OCT with a mean Spectralis/Stratus ratio of 1.58 for horizontal and vertical retinal thickness. Spectralis OCT also uses the third of the most posterior hyper-reflective line for outer boundary reference line; therefore the data might be comparable to Cirrus OCT values. Other authors comparing Stratus and Cirrus OCT reported only the mean differences in the retinal thickness (Leung et al: 20.8 μm in 35 healthy eyes7; Kiernan et al: 43 μm in 101 eyes with various macular diseases and normal eyes8; Forooghian et al: 53 μm in 33 diabetic eyes9). In a study comparing the Topcon and the Canon Optopol machines (both spectral domain technology machines) with Stratus OCT, differences were found, although these machines also use the first hyper-reflective band as posterior reference line.10
Besides different threshold algorithms and the pathology of neovascular AMD itself, differences in fixation might influence the variability of CRT and RV values measured by Stratus or Cirrus OCT (resulting in smaller values for Cirrus OCT in some patients). Recently published articles have shown that there is also a certain degree of variability between repeated measurements with Stratus OCT. We found in a previous study that differences of more than 66 μm in macular thickness program are safely pathology-related11; the findings of Patel et al were similar as they found that only differences of more than 67 μm of CRT are safely pathology-related.12 These fixation-related differences also affected the differences between Stratus and Cirrus OCT values, resulting in even lower values of CRT in Cirrus OCT in some patients. In the conversion formula this variability of the values was expressed by a residual error for CRT of 46.6 μm for Cirrus–Stratus conversion and 42.57 μm for Stratus–Cirrus conversion and by values of R2 (0.73 for CRT).
In summary, retinal thickness measurements made by Cirrus and other spectral domain technology OCT machines and Stratus OCT cannot be used interchangeably, and a change in the OCT machine used during an ongoing study should be avoided. However, to compare prior values obtained by Stratus OCT with Cirrus OCT values the formulae calculated in this study seem to be accurate.
References
Footnotes
Competing interests None declared.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Ethics Committee of the City of Vienna.