From Doheny Image Reading Center (RRP, CB, YO, ACW, SRS), Doheny Eye Institute, Keck School of Medicine of the University of Southern California, Los Angeles, California; and Smt. Kanuri Santhamma Retina Vitreous Center (RRP), L.V. Prasad Eye Institute, Hyderabad, India.
Presented in part at the Association of Research in Vision and Ophthalmology annual meeting, May 6–10, 2010, Ft. Lauderdale, Florida.
Drs. Walsh and Sadda are co-inventors of Doheny intellectual property related to optical coherence tomography that has been licensed by Topcon Medical Systems, and are members of the scientific advisory board for Heidelberg Engineering. Dr. Sadda also receives research support from Carl Zeiss Meditec, Optovue, and Optos. The remaining authors have no financial or proprietary interest in the materials presented herein.
Address correspondence to Srinivas R. Sadda, MD, Doheny Eye Institute-DEI 3623, 1450 San Pablo Street, Los Angeles, CA 90033. E-mail: firstname.lastname@example.org
Spectral-domain optical coherence tomography (SD-OCT) has been touted to offer many advantages over previous time-domain OCT instruments, such as Stratus OCT (Carl Zeiss Meditec, Dublin, CA). Chief among these advantages include improved axial resolution, higher sensitivity, and dramatically higher scanning speeds, all of which have been theorized to improve the visualization of retinal morphologic details.1
Several groups have now demonstrated the use of SD-OCT to study the effects of disease on various retinal layers.2–4 An area of great interest for many OCT researchers is the correlation of OCT morphologic features and morphometric parameters with visual function and visual outcomes following therapy. Initial work using only gross retinal thickness measurements from Stratus OCT yielded only modest correlations.5–7 Subsequently, with the advent of SD-OCT, investigators observed that the integrity of outer retinal structures, such as the external limiting membrane (ELM) and the photoreceptor inner segment/outer segment (IS/OS) junction, were of particular importance, both for identifying early disease states and for correlation with visual function.8–10
Despite the improved sensitivity and axial resolution of commercial SD-OCT instruments, these outer retinal structures are not always well seen. When these structures are not clearly visible, it is often uncertain whether this is a result of true disruption of these structures or simply insufficient image quality. Because recognizing the integrity of these structures is of clinical significance, strategies to improve their visualization are important.
An inherent property of OCT well known to degrade image quality is speckle noise artifact.11 The technique of multiple B-scan averaging was previously shown to be a useful method to decrease speckle noise in time-domain OCT,12 but this increased the time taken to perform the scan and introduced additional challenges due to eye motion artifact. However, the higher speed of SD-OCT instruments allows multiple scans to be taken quickly from the same location, particularly in SD-OCT systems that also track eye movements.13 Sakamoto et al.14 showed that the multiple B-scan averaging technique can be used to enhance SD-OCT image quality and improve visualization of the IS/OS junction and the outer nuclear layer (ONL), but they did not study the ELM. Subsequently, Spaide et al. demonstrated that multiple B-scan averaging was an important component of the their “Enhanced Depth Imaging” approach for improving visualization of the choroid.15 However, the number of scans that need to be averaged to achieve sufficient image quality for clinical assessments has not been well established.
The number of scans that can be averaged may be limited because many SD-OCT instruments do not track eye movements. We evaluated the extent of improvement in visualization of the ONL, ELM, IS/OS junction, retinal pigment epithelium (RPE), and choroid using two different commercially available averaging protocols.
OCT images were collected and reviewed from a series of consecutive patients with a variety of retinal diseases (Table 1) referred for Cirrus HD-OCT (software version 4.5, Carl Zeiss Meditec) imaging between October and November 2009 from a tertiary retina and ophthalmology subspecialty practice at the Doheny Eye Institute.
Table 1: Disease Spectrum
Averaging of the images was achieved by selective pixel profiling, which is automatically performed by the OCT machine. There was no eye tracking facility available with this machine. Selective pixel profiling averages the image data, taking into account the quality of the registration across the images that are averaged and the noise distribution in each image. In doing that, the method is best able to estimate the true signal at any given location. As such, it is a selective process and may not average across all images. Profiling is a term used to describe the act of estimating the noise distribution in each image. All OCT scans were consecutively reviewed to identify a series of patients who had at least one fovea scanned with three scan types: 1,024 × 5 high-definition B-scan raster without averaging, 1,024 × 5 with 4× averaging, and a single B-scan (1,024 × 1) with 20× averaging (passing through the foveal center). Of note, the standard image acquisition protocol for all retinal patients referred to the Imaging Unit at the Doheny Eye Institute at the time of the study included obtaining 1,024 × 5 high-definition raster B-scans with and without averaging. A few of the referring retina specialists also requested a single 20× (1,024 × 1) B-scan as part of their standard acquisition protocol. Thus, scan acquisition protocols were not specifically selected based on the presence of a particular disease or OCT finding. Approval for data collection and analysis was obtained from the institutional review board of the University of Southern California. The research adhered to the tenets set forth in the Declaration of Helsinki.
A total of 35 patients met these criteria, and one eye from each case was randomly selected for masked analysis by two trained Doheny Image Reading Center graders. Only a single B-scan, which passed through the foveal center, was evaluated for each scan type. To achieve masking to both the patient and the scan type, one grader loaded the central B-scan for viewing in random order while the second grader performed the assessments. On each graded B-scan, the following outer retinal/choroidal structures of presumed clinical value were assessed: ONL, ELM, IS/OS junction, inner RPE surface (band corresponding to interdigitation between the photoreceptor outer segments and the RPE), outer “RPE” surface (band generally considered to be the boundary between the RPE–Bruch’s complex and the choriocapillaris), and choroid (Figure).
Figure. Central foveal B-scan (1,024 A-scans) from an eye of a patient with an epiretinal membrane and pseudo-hole demonstrating the effects of B-scan averaging. (Top) Non-averaged scan (1×) showing incomplete external limiting membrane and difficulty in identifying choroidal architecture when compared with 4 times averaged (4×, Middle) and 20 times averaged (20×, Bottom) scans. The white arrow indicates the external limiting membrane seen clearly with 20× and 4×, but not with 1×. The white arrowhead marks the outer border of the choroid seen better with 20× and 4×. ONL = outer nuclear layer; IS/OS = inner segment/outer segment junction; ELM = external limiting membrane; RPE = retinal pigment epithelium.
Quality of visualization of each structure or band was evaluated based on three reading center OCT quality parameters: ability to identify the structure (ie, ability to distinguish structure from background—akin to contrast), brightness of the structure (ie, intensity), and continuity of the structure (ie, the number of A-scans in which the structure was visible). Four of the structures are thin hyperreflective bands on OCT, whereas the ONL and choroid may have significant thickness; in the case of the choroid, the full extent may not be visible. Thus, when rating the continuity and ability to identify the choroid, the graders considered the axial extent of the visible choroid when making their determinations (Figure). Each quality parameter for each structure was scored using a 0 (worst) to 3 qualitative scale (Table 2). Mean quality scores among the three scan types were computed and the chi-square test was used to examine differences between the scan types (Intercooled STATA 9 statistical software; Stata Corp, LP, College Station, TX). A significance level of .05 or less was considered to be statistically significant.
Table 2: Grading Scale
The mean quality scores for the 35 eyes of the 35 subjects included in this study are shown in Table 3, and the P values of the difference between scan types as assessed by the chi-square test are shown in Table 4. For nearly all quality parameters and structures, the quality score was higher for the averaged scans with the exception of the continuity rating for the inner RPE surface, which showed a slightly (not statistically significant different) better quality score for the non-averaged scans. The structures that showed the greatest improvement in quality score as a result of averaging were the ONL, ELM, and choroid, whereas the IS/OS junction and inner RPE surface band showed no statistically significant improvement in quality score for any quality parameter. Overall, the quality parameter that showed the greatest benefit with averaging was structure brightness, whereas continuity of the structure demonstrated the least benefit. No statistically significant improvement in quality score for continuity was observed for any structure at either averaging level. However, significant improvements in quality score were observed for brightness of the ONL, ELM, RPE outer/Bruch’s membrane band, and the choroid, and identification of the ELM and ONL. For ELM identification, 91.4% of 20× cases achieved the highest quality score (3) on all parameters compared with 82.9% of 4× cases, and only 54.3% of non-averaged cases. Quality score consistently improved with 4× averaging and improved slightly further with 20×.
Table 3: Mean Quality Score
Table 4: Chi-Square Test Significance Levels for Inter-Scan Comparisons
To evaluate whether 20× averaging conferred any benefit over 4× averaging, the difference in quality score between non-averaged and 4× averaged scans was compared with the difference between 4× and 20× averaged scans. Although no statistically significant difference was observed, the mean quality score for 20× averaged scans was higher than the 4× averaged scans for all parameters and structures. Brightness of the ONL and continuity of the ELM and IS/OS junction appeared to show the greatest benefit of increasing from 20× to 4× averaging.
This study demonstrates that multiple B-scan averaging can produce significant improvements in OCT image quality, which are relevant to several outer retinal and choroidal structures of clinical importance. In particular, B-scan averaging improved the brightness and identification of the ELM and ONL, two structures that have been shown to be of particular importance when correlating OCT morphology to visual function in a variety of retinal diseases, including non-neovascular AMD.16 In addition, visualization of the choroid and the extent or thickness of the choroid, which has recently been implicated to be of importance in several diseases, also appeared to benefit from averaging.17,18 Interestingly, quality of visualization of the IS/OS junction and the inner RPE surface (outer segment–apical RPE interdigitation band) did not improve significantly with averaging.
The current study also demonstrated that averaging appeared to have the greatest benefit on improving the brightness of structures, but little effect on increasing the continuity of these structures. However, this observation may be due in part to a ceiling effect. The structures evaluated in this study had a mean non-averaged brightness quality score of 2.02, whereas the mean non-averaged continuity score was 2.71. Thus, there was less room for improvement in the continuity score, using the existing reading center SD-OCT quality scales.
In addition, although quality scores uniformly appeared to improve further when going from 4× to 20× averaging, the difference in quality score between these two averaging levels was not statistically significant. This observation is consistent with the findings from the previous study by Sakamoto et al. that showed oversampling rates greater than 4× may not improve the image quality significantly.14 However, our study was not designed to determine the optimal averaging level to achieve the optimal clinical results because we were limited to the averaging protocols available in the OCT instrument.
The current study also has several limitations that must be considered when assessing the significance of these results. First, a subjective, coarse scale was used to assess the various quality parameters. It is possible that a finer, more granular scale may have identified more significant differences between the different scan acquisition types/averaging levels; however, inter-grader reproducibility for such a finer scale has not been established in the reading center. With future improvements in OCT segmentation algorithms for these fine structures, it is possible that a more objective, quantitative quality assessment may eventually be possible. Second, although we considered three quality parameters used in the reading center to judge OCT image quality, it is possible that there may be several other metrics that may show differential effects of B-scan averaging. Third, this study was conducted using an OCT instrument that does not track eye movements. If a patient’s eye moves during scan acquisition, scans from different retinal locations may be averaged together, thus degrading the quality of the averaged image. Such a problem may have offset some of the potential improvement in scan quality when going from 4× to 20× averaging. It is possible that further significant improvements in scan quality could be observed with higher levels of averaging in instruments that track eye movement. Fourth, this study was conducted on a single OCT instrument, the Cirrus OCT. It is not clear whether these results can be extrapolated to other SD-OCT instruments that may have differences in native non-averaged image quality. Fifth, we only considered the one A-scan density (1,024 A-scans per B-scan) available for averaging on the instrument. Our findings may not extrapolate to other scan types. Sixth, because patients were required to have been imaged with all three acquisition types to be included in the study, the sample size was relatively small. As a result, the study was underpowered to find smaller differences between groups. Finally, although we considered a variety of outer retinal structures that are relevant to visual function in various retinal disorders, inner retinal structures such as the nerve fiber layer were not assessed. The impact of averaging on identification and quantification of these inner retinal structures is unknown and may be a useful target for future study.
Multiple B-scan averaging appears to be a simple technique by which SD-OCT B-scan image quality can be improved. The benefits appear to be greatest for visualization of the ELM, ONL, and choroid. When designing studies requiring assessments of these structures, investigators should consider employing scan acquisition protocols that use B-scan averaging.
- Sayanagi K, Sharma S, Yamamoto T, Kaiser PK. Comparison of spectral-domain versus time-domain optical coherence tomography in management of age-related macular degeneration with ranibizumab. Ophthalmology. 2009;116:947–955. doi:10.1016/j.ophtha.2008.11.002 [CrossRef]
- Mitamura Y, Aizawa S, Baba T, Hagiwara A, Yamamoto S. Correlation between retinal sensitivity and photoreceptor inner/outer segment junction in patients with retinitis pigmentosa. Br J Ophthalmol. 2009;93:126–127. doi:10.1136/bjo.2008.141127 [CrossRef]
- Vasconcelos-Santos DV, Sohn EH, Sadda S, Rao NA. Retinal pigment epithelial changes in chronic Vogt-Koyanagi-Harada disease: fundus autofluorescence and spectral domain-optical coherence tomography findings. Retina. 2010;30:33–41. doi:10.1097/IAE.0b013e3181c5970d [CrossRef]
- Witkin AJ, Ko TH, Fujimoto JG, et al. Ultra-high resolution optical coherence tomography assessment of photoreceptors in retinitis pigmentosa and related diseases. Am J Ophthalmol. 2006;142:945–952. doi:10.1016/j.ajo.2006.07.024 [CrossRef]
- Scott IU, VanVeldhuisen PC, Oden NL, et al. SCORE Study report 1: baseline associations between central retinal thickness and visual acuity in patients with retinal vein occlusion. Ophthalmology. 2009;116:504–512. doi:10.1016/j.ophtha.2008.10.017 [CrossRef]
- Keane PA, Liakopoulos S, Chang KT, et al. Relationship between optical coherence tomography retinal parameters and visual acuity in neovascular age-related macular degeneration. Ophthalmology. 2008;115:2206–2214. doi:10.1016/j.ophtha.2008.08.016 [CrossRef]
- Browning DJ, Apte RS, Bressler SB, et al. Association of the extent of diabetic macular edema as assessed by optical coherence tomography with visual acuity and retinal outcome variables. Retina. 2009;29:300–305. doi:10.1097/IAE.0b013e318194995d [CrossRef]
- Bearelly S, Chau FY, Koreishi A, Stinnett SS, Izatt JA, Toth CA. Spectral domain optical coherence tomography imaging of geographic atrophy margins. Ophthalmology. 2009;116:1762–1769. doi:10.1016/j.ophtha.2009.04.015 [CrossRef]
- Brar M, Kozak I, Cheng L, et al. Correlation between spectral-domain optical coherence tomography and fundus autofluorescence at the margins of geographic atrophy. Am J Ophthalmol. 2009;148:439–444. doi:10.1016/j.ajo.2009.04.022 [CrossRef]
- Fleckenstein M, Charbel Issa P, Helb HM, et al. High-resolution spectral domain-OCT imaging in geographic atrophy associated with age-related macular degeneration. Invest Ophthalmol Vis Sci. 2008;49:4137–4144. doi:10.1167/iovs.08-1967 [CrossRef]
- Hangai M, Yamamoto M, Sakamoto A, Yoshimura N. Ultrahigh-resolution versus speckle noise-reduction in spectral-domain optical coherence tomography. Opt Express. 2009;17:4221–4235. doi:10.1364/OE.17.004221 [CrossRef]
- Sander B, Larsen M, Thrane L, Hougaard JL, Jorgensen TM. Enhanced optical coherence tomography imaging by multiple scan averaging. Br J Ophthalmol. 2005;89:207–212. doi:10.1136/bjo.2004.045989 [CrossRef]
- Hammer DX, Ferguson RD, Magill JC, et al. Active retinal tracker for clinical optical coherence tomography systems. J Biomed Opt. 2005;10:024038. doi:10.1117/1.1896967 [CrossRef]
- Sakamoto A, Hangai M, Yoshimura N. Spectral-domain optical coherence tomography with multiple B-scan averaging for enhanced imaging of retinal diseases. Ophthalmology. 2008;115:1071–1078. doi:10.1016/j.ophtha.2007.09.001 [CrossRef]
- Spaide RF, Koizumi H, Pozzoni MC. Enhanced depth imaging spectral-domain optical coherence tomography. Am J Ophthalmol. 2008;146:496–500. doi:10.1016/j.ajo.2008.05.032 [CrossRef]
- Pappuru RR, Ouyang Y, Nittala MG, et al. Relationship between outer retinal substructures and visual acuity in eyes with dry age-related macular degeneration [published online ahead of print June 17, 2011]. Invest Ophthalmol Vis Sci.
- Ikuno Y, Tano Y. Retinal and choroidal biometry in highly myopic eyes with spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2009;50:3876–3880. doi:10.1167/iovs.08-3325 [CrossRef]
- Imamura Y, Fujiwara T, Margolis R, Spaide RF. Enhanced depth imaging optical coherence tomography of the choroid in central serous chorioretinopathy. Retina. 2009;29:1469–1473. doi:10.1097/IAE.0b013e3181be0a83 [CrossRef]
|Choroidal neovascular membrane|
|Dry age-related macular degeneration|
|Branch retinal vein occlusion|
|Central retinal vein occlusion|
|Proliferative diabetic retinopathy|
|Clinically significant macular edema|
|Cystoid macular edema|
|0||Cannot differentiate at all from background||Not visible||Not visible in any A-scan|
|1||Can differentiate in up to 1/3 of B-scan||Low||Seen in up to 33% of A-scans|
|2||Can differentiate in 1/3 to 2/3 of B-scan||Medium||Seen in 34% to 66% of A-scans|
|3||Can differentiate in 2/3 or more of B-scan||Bright||Seen in 67% or more of A-scans|
Mean Quality Score
Chi-Square Test Significance Levels for Inter-Scan Comparisonsa
|NA vs 4× vs 20×||NA vs 4×||4× vs 20×||NA vs 20×||NA vs 4× vs 20×||NA vs 4×||4× vs 20×||NA vs 20×||NA vs 4× vs 20×||NA vs 4×||4× vs 20×||NA vs 20×|