Factors affecting signal strength in spectral-domain optical coherence tomography
Abstract
Purpose
To identify ocular factors that affect signal strength in spectral-domain optical coherence tomography (SD-OCT).
Methods
Data from 1312 participants of the population-based Singapore Malay Eye Study-2 (SiMES-2) were included in the analysis. All participants underwent standardized ophthalmic examination, including measurements of best-corrected visual acuity (BCVA), refractive error, axial length, corneal curvature and presence of cataracts. Optic disc and macular cube scans were acquired using the Cirrus HD-OCT (software version 6.0, Carl Zeiss Meditec, Dublin, CA, USA). Signal strength of the optical coherence tomography (OCT) scan was recorded for each study eye. Multivariable linear regression analyses were performed to evaluate the associations between ocular factors and signal strength of the OCT scans.
Results
: The mean (±SD) age of our study participants was 61 ± 9 years, and 44.6% were male. Mean optic disc scan signal strength was 7.90 ± 1.25, range = 0–10, while mean macular scan signal strength was 8.80 ± 1.27, range = 0–10. In multivariable regression analyses, poorer signal strength in optic disc and macular cube scans was each associated with older age (per decade, β = −0.373, p < 0.001; β = −0.373, p < 0.001, respectively), poorer BCVA (per logMAR line; β = −0.123, p < 0.001; β = −0.156, p < 0.001, respectively), greater degree of myopia (per negative dioptre of spherical equivalent; β = −0.112, p < 0.001; β = −0.117, p < 0.001, respectively), presence of cortical cataracts (β = −0.331, p < 0.001; β = −0.314, p < 0.001, respectively) and presence of posterior subcapsular cataracts (β = −0.910, p < 0.001; β = −0.797, p < 0.001, respectively).
Conclusion
We found that older age, poorer BCVA, greater degree of myopia and presence of cortical and posterior subcapsular cataracts were associated with reduced signal strength in Cirrus SD-OCT. Our findings provide information on the barriers to obtaining good image quality when using SD-OCT, and allow clinicians to potentially identify individuals who are more likely to have unreliable OCT measurements.
Introduction
Spectral-domain optical coherence tomography (SD-OCT) is an advanced imaging technology increasingly used in ophthalmology as a diagnostic and monitoring tool for glaucoma and retinal diseases (Mwanza et al. 2013; Browning et al. 2004; Tanner et al. 2001). This technology uses automated software algorithms to quantify retinal and optic nerve head parameters through segmentation of intraretinal layers.
Optic disc cube scans in SD-OCT allow accurate measurement of peripapillary retinal nerve fibre layer (RNFL) thickness, enabling early diagnosis of glaucoma and monitoring of glaucoma progression (Kaushik et al. 2014). On the other hand, macular cube scans in SD-OCT allows measurement of macular thickness, which allow objective assessment of macular changes such as resolution of oedema after treatment (Mushtaq et al. 2014). However, the performance of the OCT layer segmentation algorithms can be affected by poor image quality, leading to erroneous demarcation of the retinal layers and inaccurate measurements, and thus delay in early detection or misdiagnosis of disease. Previous studies have shown that poorer image quality was associated with underestimation of RNFL and macular thickness (Cheung et al. 2008; Vizzeri et al. 2009; Samarawickrama et al. 2010; Cheung et al. 2012). Differences in SD-OCT scan quality was found to be the strongest factor associated with interscan variability of tissue thickness measurements (Vizzeri et al. 2009). Furthermore, it was also previously reported that measurements of optic disc morphology such as cup–disc ratio and rim area are underestimated in scans with poorer quality (Samarawickrama et al. 2010).
Manufacturers of OCT machines provide parameters to indicate the image quality of OCT scans. For the Cirrus High Definition SD-OCT (Carl Zeiss Meditec, Dublin, CA, USA), each OCT scan is indicated with a signal strength parameter, which ranges from 0 to 10. The manufacturer defines scans of adequate quality to be those of signal strength 6 or above (Carl Zeiss Meditec Inc, 2008). Scans of signal strength 5 and below would produce RNFL thickness measurements with greater variability attributable to inaccurate segmentation (Balasubramanian et al. 2009). Previous studies found that various ocular factors such as corneal dryness, cataracts, astigmatism and pupil size affect scan quality and signal strength in OCT scans (Stein et al. 2006; Smith et al. 2007; van et al. 2006; Hwang et al. 2012). However, most of these studies were performed using time-domain OCT. As SD-OCT is the more commonly used imaging domain in current clinical practice, it is important to examine whether these factors are similarly relevant in higher definition SD-OCT. Furthermore, previous studies only performed analyses at the univariate level, without taking into account potential confounders. Thus, findings from previous studies may not be entirely conclusive.
The aim of this study was to examine the effects of various ocular factors on signal strength obtained by the Cirrus SD-OCT in a population-based study. Findings in this study will improve our understanding on the barriers to obtaining optimal signal strength for SD-OCT.
Materials and Methods
Data for this analysis was obtained from the SiMES-2, a population-based cohort study of eye diseases in Malay adults aged 46–86 years living in Singapore (Rosman et al. 2012). Written informed consent was obtained from all participants. The study adhered to the tenets of the Declaration of Helsinki, and ethics committee approval was obtained from the SingHealth Centralised Institutional Review Board. All participants underwent standardized and comprehensive ophthalmic examinations. Participants were included in the study if SD-OCT imaging was performed in at least one eye. If both eyes were imaged, the eye with the poorer BCVA was included in the analysis. Eyes which did not have signal strength data available or had missing data in any of the variables involved in the linear regression model were excluded.
SD-OCT imaging
An experienced operator performed the SD-OCT scans after pupil dilation using tropicamide 1% and phenylephrine hydrochloride 2.5%. The Cirrus HD-OCT (software version 6.0, Carl Zeiss Meditec Inc.) is a commercially available SD-OCT with a scan speed of 27 000 axial scans per second and an axial resolution of 5 μm. We acquired optic disc and macular scans using the 200 × 200 optic disc cube protocol and 512 × 128 macular cube scan protocol, respectively. Each scan captured an area of 6 × 6 mm2. During image acquisition, the subject's pupil first was centred and focused in the iris viewport, and the line-scanning ophthalmoscope with ‘auto focus’ mode then was used to optimize the view of the retina. The ‘centre’ and ‘enhance’ modes were used to optimize the Z-offset and scan polarization, respectively, for the OCT scan to maximize the OCT signal. The built-in software generated a signal strength score for each scan taken, which was recorded.
Measurement of ocular factors
All participants underwent a standardized ophthalmic examination which includes BCVA testing using a logarithmic minimal angle of resolution (logMAR) number chart, static refraction, subjective refraction, axial length measurement, slit-lamp biomicroscopy and dilated fundus examination. Static refraction and corneal curvature of each eye was measured using an autorefractor and keratometer (Canon RK 5 Auto Ref-Keratometer; Canon, Inc., Ltd., Tochigiken, Japan), and the mean of five measurements used for analysis. Spherical equivalent refraction was calculated as the sum of the value of the spherical value and half of the cylindrical value. Axial length was measured with a non-contact partial coherence laser interferometry (IOLMaster Version 3.01; Carl Zeiss Meditec AG, Jena, Germany), and the mean of five measurements was used for the analysis. Lens opacity was assessed using the Lens Opacities Classification System (LOCS) III with a Haag-Streit slit-lamp microscope (model BQ-900) in comparison with standard photographic slides for nuclear opalescence, nuclear colour, cortical and posterior subcapsular cataract (Chylack et al. 1993). Detailed methodology on lens opacity grading has been reported elsewhere (Chylack et al. 1993). Briefly, nuclear opalescence and colour were graded in comparison with six standard slit-lamp images of increasing cataract severity, while cortical cataracts and posterior subcapsular cataracts were each graded in comparison with five retroillumination images of the respective cataract type. Presence of nuclear cataract was defined as nuclear opalescence or nuclear colour grading of 4 or above. Presence of cortical cataract was defined as cortical grading of 2 or above, while posterior subcapsular cataract was defined as posterior subcapsular grading of 2 or above.
Statistical analysis
Statistical analysis was performed using spss Version 20.0 (SPSS, Inc., Chicago, IL, USA). We performed univariate and multivariable analyses to determine the association between ocular factors (independent variables) and OCT signal strength (dependent variable). Only ocular parameters with p < 0.05, found in univariate analysis, were included in the multivariable linear regression analysis model. In the linear regression models, least squares fitting was applied.
Results
Of the 1901 participants in SiMES-2, 589 did not undergo SD-OCT imaging and were excluded, leaving 1312 participants to be included in this study. The most frequent reason for not undergoing SD-OCT imaging was the participant's inability to fixate during the scan. The mean age of our study participants was 60.8 ± 9.0 years, 44.6% of whom were male. Mean optic disc scan signal strength was 7.90 (SD = 1.25, range = 0–10), while mean macular scan signal strength was 8.80 (SD = 1.27, range = 0–10).
Table 1 compares the demographics and ocular characteristics between the included and excluded participants. In general, excluded participants were older, had poorer BCVA, greater degree of astigmatism and were more likely to have cataracts (all p < 0.001).
| Parameter | Included (N = 1312) | Excluded (N = 589) | p value |
|---|---|---|---|
| Age | 60.8 ± 9.0 | 69.6 ± 10.4 | <0.001 |
| Male gender, N (%) | 585 (44.6) | 279 (47.4) | 0.273 |
| Best-corrected visual acuity, logMAR | 0.199 ± 0.171 | 0.418 ± 0.464 | <0.001 |
| Axial length, mm | 23.612 ± 1.106 | 23.638 ± 1.251 | 0.690 |
| Cylindrical refractive error, D | 0.981 ± 0.702 | 1.291 ± 0.957 | <0.001 |
| Corneal curvature, mm | 7.652 ± 0.252 | 7.646 ± 0.270 | 0.678 |
| Presence of nuclear cataract, N (%) | 497 (37.9) | 299 (50.8) | <0.001 |
| Presence of cortical cataract, N (%) | 393 (30.0) | 224 (38.0) | 0.001 |
| Presence of posterior subcapsular cataract, N (%) | 58 (4.4) | 93 (15.8) | <0.001 |
- logMAR = logarithmic minimum angle of resolution, D = dioptre.
- Variables are presented as mean ± SD, or N (%) as appropriate.
Table 2 shows the association analyses between ocular factors and signal strength from optic disc cube scan. In multivariable analysis, older age (per decade, β = −0.373, p < 0.001), poorer BCVA (per logMAR line, β = −0.123, p < 0.001), longer axial length (per mm, β = −0.234, p < 0.001), greater degree of cylindrical refractive error (per dioptre, β = −0.160, p < 0.001), presence of cortical cataracts (β = −0.331, p < 0.001) and presence of posterior subcapsular cataracts (β = −0.910, p < 0.001) were associated with poorer signal strength in optic disc cube scans. In addition, higher spherical power (per negative dioptre, β = −0.105, p < 0.001) and higher spherical equivalent (per negative dioptre, β = −0.112, p < 0.001) were also associated with poorer signal strength. Amongst the significant factors, age and axial length had the strongest effect on optic disc cube scan signal strength (standardized β = −0.265 and −0.207, respectively).
| Variable | Univariate analysis | Multivariable analysisa | |||||
|---|---|---|---|---|---|---|---|
| β (95% CI) | Standardized β | p | β (95% CI) | Standardized β | p | Partial R2 | |
| Age, per decade increase | −0.561 (−0.631 to −0.491) | −0.399 | <0.001 | −0.373 (−0.450 to −0.296) | −0.265 | <0.001 | 0.050 |
| Gender (Female) | 0.106 (−0.031 to 0.243) | 0.042 | 0.128 | 0.046 (−0.074 to 0.165) | 0.018 | 0.454 | 0.000 |
| Best-corrected visual acuity, logMAR lines | −0.273 (−0.311 to −0.234) | −0.363 | <0.001 | −0.123 (−0.162 to −0.083) | −0.164 | <0.001 | 0.021 |
| Axial length, mm | −0.167 (−0.228 to −0.106) | −0.148 | <0.001 | −0.234b (−0.295 to −0.172) | −0.207 | <0.001 | 0.031 |
| Spherical refractive error, per negative D | −0.072 (−0.103 to −0.041) | −0.126 | <0.001 | −0.105c (−0.133 to −0.078) | −0.185 | <0.001 | 0.030 |
| Cylindrical refractive error, per negativ D | −0.431 (−0.525 to −0.336) | −0.242 | <0.001 | −0.160 (−0.247 to −0.073) | −0.090 | <0.001 | 0.007 |
| Corneal curvature, mm | −0.274 (−0.545 to −0.002) | −0.055 | 0.048 | 0.110 (−0.158 to 0.378) | 0.022 | 0.421 | 0.000 |
| Presence of nuclear cataract | −0.479 (−0.617 to −0.341) | −0.186 | <0.001 | −0.113 (−0.238 to 0.012) | −0.044 | 0.076 | 0.002 |
| Presence of cortical cataract | −0.717 (−0.861 to −0.573) | −0.263 | <0.001 | −0.331 (−0.466 to −0.197) | −0.121 | <0.001 | 0.013 |
| Presene of posterior subcapsular cataract | −1.566 (−1.898 to −1.233) | −0.249 | <0.001 | −0.910 (−1.212 to −0.607) | −0.145 | <0.001 | 0.019 |
- CI = confidence interval, logMAR = logarithmic minimum angle of resolution, D = dioptre.
- a Age, gender and variables with a p value <0.05 in univariate models were included in the multivariable model.
- b Spherical refractive error was excluded from the model because of collinearity with spherical refractive error.
- c Axial length was excluded from the model because of collinearity with spherical refractive error.
Similar results were observed in macular cube scans (Table 3). Older age (per decade, β = −0.373, p < 0.001), poorer BCVA (per logMAR line, β = −0.156, p < 0.001), longer axial length (per mm, β = −0.211, p < 0.001), greater degree of cylindrical refractive error (per dioptre, β = −0.126, p < 0.001), presence of cortical cataracts (β = −0.314, p < 0.001) and presence of posterior subcapsular cataracts (β = −0.797, p < 0.001) were associated with poorer signal strength in optic disc cube scans. In addition, higher spherical power (per negative dioptre, β = −0.112, p < 0.001) and higher spherical equivalent (per negative dioptre, β = −0.117, p < 0.001) were also associated with poorer signal strength. Amongst the significant factors, age and BCVA had the strongest effect on macular cube scan signal strength (standardized β = −0.265 and −0.213, respectively).
| Variable | Univariate analysis | Multivariable analysisa | |||||
|---|---|---|---|---|---|---|---|
| β (95% CI) | Standardized β | p | β (95% CI) | Standardized β | p | Partial R2 | |
| Age, per decade increase | −0.567 (−0.637 to −0.497) | −0.402 | <0.001 | −0.373 (−0.450 to −0.296) | −0.265 | <0.001 | 0.049 |
| Gender (Female) | 0.021 (−0.117 to 0.159) | 0.008 | 0.764 | −0.051 (−0.171 to 0.069) | −0.020 | 0.403 | 0.000 |
| Best-corrected visual acuity, logMAR lines | −0.294 (−0.330 to −0.258) | −0.401 | <0.001 | −0.156 (−0.194 to −0.118) | −0.213 | <0.001 | 0.035 |
| Axial length, mm | −0.172 (−0.233 to −0.110) | −0.150 | <0.001 | −0.211b (−0.272 to −0.150) | −0.185 | <0.001 | 0.025 |
| Spherical refractive error, per negative D | −0.080 (−0.111 to −0.049) | −0.139 | <0.001 | −0.112c (−0.139 to −0.084) | −0.195 | <0.001 | 0.034 |
| Cylindrical refractive error, per negative D | −0.398 (−0.492 to −0.304) | −0.224 | <0.001 | −0.126 (−0.212 to −0.040) | −0.071 | 0.004 | 0.004 |
| Corneal curvature, mm | −0.479 (−0.749 to −0.209) | −0.096 | 0.001 | −0.150 (−0.415 to 0.115) | −0.030 | 0.268 | 0.001 |
| Presence of nuclear cataract | −0.407 (−0.547 to −0.267) | −0.156 | <0.001 | −0.020 (−0.145 to 0.104) | −0.008 | 0.750 | 0.000 |
| Presence of cortical cataract | −0.714 (−0.859 to −0.570) | −0.259 | <0.001 | −0.314 (−0.448 to −0.180) | −0.114 | <0.001 | 0.011 |
| Presence of posterior subcapsular cataract | −1.510 (−1.831 to −1.189) | −0.247 | <0.001 | −0.797 (−1.089 to −0.504) | −0.130 | <0.001 | 0.015 |
- CI = confidence interval, logMAR = logarithmic minimum angle of resolution, D = dioptre.
- a Age, gender and variables with a p value <0.05 in univariate models were included in the multivariable model.
- b Spherical refractive error was excluded from the model because of collinearity with spherical refractive error.
- c Axial length was excluded from the model because of collinearity with spherical refractive error.
Discussion
In this study, we examined the effects of various ocular factors on signal strength of Cirrus SD-OCT scans. We found that poorer BCVA, greater degree of myopia and astigmatism and presence of cortical and posterior subcapsular cataracts to be significantly associated with reduced signal strength. This is the first study which investigated the effects of ocular factors on signal strength in SD-OCT using a population-based sample. The findings of this study provide better understanding on the barriers to obtaining high-quality OCT images and allow clinicians to potentially identify individuals who are more likely to have less reliable OCT measurements.
Signal strength is an important consideration when assessing results of OCT imaging. Previous studies, performed using Cirrus OCT, found that even when signal strength fell within the range considered acceptable to the manufacturer (signal strength ≥6), significant interscan variability in tissue thickness measurements was still observed, with measurements being underestimated with lower signal strength (Cheung et al. 2012; Balasubramanian et al. 2009). In a previous population-based study, it was found that poorer signal strength was associated with thinner RNFL thickness measurements, indicating that underestimation of parameter thickness may occur in SD-OCT scans with poor signals (Cheung et al. 2011). Hence, images should be taken at the maximum signal strength possible, to ensure reliable measurements are obtained. In addition, this also implies that identification of factors that affect signal strength is essential.
Consistent with our study, Na et al. (2012) investigated a small hospital-based sample (n = 92) and found that older age and poorer visual acuity (VA) were associated with poorer signal strength. This may be because older individuals and individuals with poorer VA may have poorer fixation during OCT scan acquisition, leading to more motion artefacts and thus compromised signal strength.
On the other hand, we found that myopia and astigmatism were independently associated with poorer signal strength. Similarly, Hwang et al. (2012) investigated the effects of astigmatism induced by toric contact lenses on signal strength in Cirrus SD-OCT, and found that three dioptres of induced astigmatism corresponded to a decline in signal strength of 1.4 on average. In our univariate analysis, a similar estimate was found; three dioptres of astigmatism corresponded to a decline in signal strength of 1.2. However, on multivariable analysis, the independent effect of three dioptres of astigmatism corresponded to a decline of signal strength of only 0.4. In addition, compared to spherical refractive error, we observed that astigmatism had a larger effect on signal strength. Based on the effect sizes obtained from our models, in both optic disc cube and macular cube scans, myopia of -5D corresponded to an estimated decline of signal strength by 0.5 unit. In comparison, in both optic disc cube scans and macular cube scans, increase in degree of astigmatism of 3D and 4D, respectively, corresponded to a decline of signal strength by 0.5 unit. Taken together, these findings indicate that signal strength from SD-OCT scans is rather robust against refractive errors, with signal strength only substantially affected in the presence of very high myopia or high astigmatism.
In our study, posterior subcapsular and cortical cataracts were significantly associated with decreased signal strength in both optic disc cube and macular cube scans. In contrast, the presence of nuclear cataracts did not have a statistically significant effect on signal strength. Coincidentally, Velthoven et al. (2006) also reported that signal strength in eyes with predominantly nuclear cataracts tend to be higher than in eyes with predominantly posterior subcapsular or cortical cataracts. Overall, this suggests that removal of the posterior subcapsular and cortical cataract types may more likely lead to more significantly improved signal strength in SD-OCT scans, compared to removal of nuclear cataract types. In addition, this information also indicates that OCT scans should be repeated in patients after they undergo cataract removal surgery, and this is especially relevant amongst individuals with history of previous posterior subcapsular or cortical cataracts.
The strengths of this study include its large sample size. Furthermore, our study included comprehensive measurement of potential confounding factors which were taken into account in multivariable analysis. However, this study has limitations. The range of OCT scan signal strength obtained in our samples predominantly covers the range of 4–10, with few OCT scans with signal strength of three or below. Thus, our findings may have limited generalizability to scans of signal strength 3 or below. A few ocular factors that were previously reported to affect signal strength, such as corneal dryness and pupil size, were not included in our analysis due to the lack of availability of these data. It should also be highlighted that our findings are limited to the Cirrus OCT, and may not be entirely generalized to other commercially-available SD-OCT modalities.
In conclusion, we found that older age, poorer BCVA, greater myopia and presence of cortical and posterior subcapsular cataracts were associated with reduced signal strength in SD-OCT. This information is important in understanding the barriers to obtaining high-quality SD-OCT scans. This may also allow clinicians to better identify individuals who are more likely to have less reliable OCT measurements. Our results also show that SD-OCT signal strength is minimally affected by ocular factors; hence, reliable scans should be attainable from most patients, even those with nuclear cataracts or relatively severe refractive errors.
