Clinical adoption of spectral-domain optical coherence tomography (SD-OCT) enables high-resolution visualization of chorioretinal pathologies affecting retinal layers in the macula.1 It is a sensitive tool for noninvasive diagnosis, monitoring of disease progression, and assessing response to therapeutic interventions.2–5 Computation of quantitative parameters from SD-OCT, such as total retinal, choroidal, nerve fiber, and ganglion cell layer thicknesses, have been investigated in a wide range of diseases; these include age-related maculopathy, central serous retinopathy, polypoidal choroidal vasculopathy, multiple sclerosis, and glaucoma.6–9 Normal variations of the macular retinal thickness have also been identified, along with correlations with age, gender, degree of myopia, smoking, and BMI.10
More specifically in the outer retinal layers, changes in the photoreceptor outer segment (PROS) thickness and volume have been identified in diseased states. Qualitatively, pathological changes in photoreceptor structure can be identified using SD-OCT. However, more subtle variations of the macular PROS structure — only reliably quantified with software processing — have also been identified using SD-OCT. For instance, it has been shown that PROS volume is decreased in patients with birdshot chorioretinopathy (BSCR); furthermore, those with abnormal electroretinograms (ERGs) have lower PROS volume than those with normal ERGs.11 In addition, this decreased PROS volume in BSCR can increase following systemic immunomodulatory treatment.12 In panuveitis secondary to Vogt-Koyanagi-Harada (VKH) syndrome, increases in PROS volume parallel improvement in visual acuity following resolution of serous macular detachments.13 In diabetic macular edema (DME), the PROS thickness was negatively correlated with visual acuity.14 In a study of patients with dry age-related macular degeneration (AMD), the thickness and volume of the PROS within the central foveal subfield was negatively correlated with visual acuity.15 Unfortunately, there are only a few studies that have characterized the variation of outer retinal layers in healthy subjects for establishing a normative baseline.16
The PROS structure on the SD-OCT can be quantified using widely available tools in a clinical setting. Therefore, knowledge of its normal anatomical variation can be helpful in guiding diagnosis and treatment decisions. The current study aims to evaluate the correlation of PROS volume with age, gender, refractive error, and presence of vitreomacular adhesion (VMA) in a cross-section of healthy subjects. Segmentation of the layer and volume calculations were performed from SD-OCT using a validated automated algorithm.17 The results are discussed in the context of the latest literature on PROS changes in healthy and pathological states.
Patients and Methods
In this prospective study, we enrolled the fellow normal eye of patients being assessed for a unilateral ocular disease (posterior vitreous detachment and cataract). SD-OCT scans were obtained using the Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA). The scan was performed using the 512 × 128 scan pattern, and only one attempt was made at scanning each eye. Each study eye was pharmacologically dilated prior to SD-OCT scanning. Exclusion criteria included poor signal strength (< 6). Charts of all patients were reviewed and demographic information (age and gender) was gathered. As a surrogate for axial length, we used refractive error (converted to spherical equivalents). The presence of VMA was assessed through detailed examination of all SD-OCT B-scans by a medical retina specialist (FF). Ethics approval was obtained from the Research Ethics Board at the University of British Columbia and the research study adhered to the tenets set forth in the Declaration of Helsinki.
A total of 73 SD-OCT images (from 73 patients) were analyzed, each with 128 B-scans at 0.0472 μm distance apart. There were 1,024 A-scans per B-scan, at a scale of 0.002 μm/pixel. Each A-scan comprised of 512 pixels, at a scale of 0.0117 μm/pixel. Calculating the PROS volume required delineating the inner segment/outer segment junction (the ellipsoid layer) and the apical retinal pigment epithelium (RPE) boundary. This was accomplished by an automatic segmentation algorithm, developed originally by Lang et al., which has been validated to be among the most accurate methods currently available.18 This algorithm belongs to the subset of machine learning methods that uses a trained classifier (more specifically, the random forest classifier) to first produce a probability map of retinal boundary layers.17 Subsequently, the boundaries of as many as nine retinal layers can be determined by a refinement step based on the optimal graph-based search algorithm. To calculate the PROS volume, the region of interest was confined to a circular area of 6 mm centered on the fovea. The volume of the disc was calculated as a three-dimensional (3-D) geometric surface, generated from the segmentation boundaries between the PROS and the inner RPE. All image processing was performed in the MATLAB software (MathWorks, Natick, MA).
For statistical analysis of the results, a univariate general linear model was used with the PROS volume as the scale linear response variable. In this healthy population, the independent effects of interest were age, spherical equivalent, gender, and the presence of VMA. Image signal quality, rated on a scale of 10, was also included as a covariate as it could potentially affect the accuracy of automatic segmentations. The first three variables were modeled as numerical covariates, and the latter two as categorical factors. Only the primary effects were included in the linear model, without higher-order interactions. The significance of the model parameters was tested with alpha set at a P value of less than .05. The SPSS software (IBM SPSS Statistics v.19; SPSS, Chicago, IL) was used for processing and compiling the statistical analysis. Since the measurements were computed automatically, no intraclass correlation statistics were performed.
An example of a B-scan with the PROS segmentation results is shown in the Figure. Of the 73 OCT images analyzed by the retinal layer segmentation algorithm, five were excluded as the automated processing either failed to complete, or could only be completed with manual manipulation at the cost of introducing operator variability. The cause in all these excluded samples was related to scan acquisition alignment issues such that not all B-scans covered the region of interest.
Representative example of results of the automatic segmentation of the inner segment/outer segment junction and inner boundary of the retinal pigment epithelium, which define the photoreceptor outer segment volumetric surface.
The response variable was normally distributed according to the Shapiro-Wilk test. The standardized residuals of the model were also normally distributed. Within the 68 included subjects, the gender distribution was 54% female and 46% male. The average age was 51.6 years ± 17.7 years (range: 13 years to 85 years), and VMA was present in 72% of the cases. The mean spherical equivalent was −1.02 ± 2.69. The mean PROS volume was 0.81 mm3 ± 0.09 mm3. The mean signal strength was 9.02 ± 1.05.
The general linear model results for PROS volume are summarized in the Table. The only significant parameter was age, with a positive effect size. In other words, PROS volume increased with age. The effect of gender, though not statistically significant with a P value of .089, had the largest effects size, with females having a larger PROS volume in this sample. The spherical equivalent and presence of VMA were largely insignificant. The effect of signal quality is shown to be a significant factor.
General Linear Model Analysis for Photoreceptor Outer Segment Volume
Quantitative knowledge of normal anatomical variation of retinal layers is essential for recognition of pathological states. The total thickness of the macular retina on SD-OCT scans has been widely reported, most notably in a study of 67,321 normal subjects in UK;10 It was demonstrated that in the central macula, the thickness was positively correlated with age, whereas in the other macular subfields, it was negatively correlated with age. It was also positively correlated with female gender, myopia, smoking, and BMI. In a study of young Chinese patients between the ages of 18 years and 30 years, there was no correlation with age, but there were significant correlations with gender, axial length, and spherical equivalent.19 A systematic review of the literature has summarized that overall macular thickness and volume decrease with age, mostly due to thinning of the nerve fiber layer and inner retinal layers.16
This prospective study focused on the variation of PROS volume. The single patient factor that had a statistically significant effect was age. The influence of VMA has not been previously reported in the literature; it was found to be non-significant in this study. The effect of gender and axial length on PROS thickness has been previously reported, as discussed below. We found that both were not statistically significant in our sample of healthy subjects, although gender had a large effect size.
This study corroborates the small number of other observational studies on changes in the PROS volume with age.16,20,21 However, the exact physiological mechanism for this has not been elucidated. Ooto et al. speculate that this is due to decreased RPE phagocytosis of discs shed at the distal outer segment of photoreceptors.20 However, no histological nor in vitro evidence of this proposed mechanism has been reported. In fact, in vitro experimental evidence from animal cells suggests that the phagocytic activity of aged RPE cells is well maintained; according to Chen et al., with aging the RPE layer remains uninterrupted, even as the number of cells decreases.22 This leads to an increase in the number of photoreceptors supported by each RPE cell, and hence higher metabolic demands; the RPE cells respond to this stress by giant cell formation and multinucleation which compensate to meet the metabolic and homeostatic needs of the photoreceptors.22 More investigation is needed to determine if the RPE giant cells can maintain their phagocytic function in vivo, and the precise mechanism of increased PROS volume with age.
There are few reports of PROS thickness and volume changes in absence of chorioretinal pathology. In a study of 256 Japanese subjects, it was found that the PROS thickness correlated positively with age in women only after adjusting for axial length.20 However, there is ambiguity with regard to the exact definition of the PROS segmentation in this study, which Demirkaya et al.23 suggest may have overestimated the thickness measurements. A Dutch study with 120 subjects found no correlation of PROS thickness with age, after adjusting for spherical equivalent, image quality, and gender.23 In an analysis of variance test of 100 subjects, the PROS thickness positively correlated with axial length and degree of myopia independently, but age was not included in the statistical analysis.21 In this study, we did not find that gender or refractive error significantly correlated with PROS volume.
In the above studies on the shape and distribution of the PROS, there is limited control of relevant fixed factors and covariates in the statistical analysis. Our study is the first to investigate the correlation of the PROS 3-D volume within the macula while adjusting for the effect of age, gender, refractive error, and vitreomacular adhesion, as well as SD-OCT signal quality. As summarized in the introduction, in diseased states such as BSCR, DME, and dry AMD, the PROS thickness or volume decreases. Therefore, we propose that an increase in PROS volume on SD-OCT is the expected normal physiological change with increasing age.
A final notable discussion is that the signal quality was inversely proportional to PROS volume to a significant degree. The mean signal quality of all images was above 9 out of a maximum possible value of 10. From the total 73 images originally included, only four had a signal quality of 6, which was the lowest permissible signal quality for a valid scan. This sample size is insufficient to conclude the true impact of signal quality. It has been previously reported that the accuracy of retinal layer segmentations is dependent on signal strength, mainly in context of measuring nerve fiber layer thickness.24 In our study, increased noise in the image can cause blurring of the hyperreflective signals representing the ellipsoid zone and the outer segment interdigitation at the RPE apical surface. It is plausible that this can lead to an overestimation error by the automatic segmentation algorithm. However, a study including more image samples with low signal strength are required to determine if this error effect is random or systematic.
Among the limitations of this study is that the spherical equivalent was used as a surrogate for the ocular axial length. This measurement possibly confounds other sources of refractive error along with axial length, which may not have physiological bearing on changes to the PROS volume. Another limitation is that reproducibility of the findings are contingent on high-resolution and high-quality scans, using the same segmentation method which uses a training set based on similarly high quality data. Furthermore, the automated algorithm has not been optimized for use in disease states and has a high probability of segmentation failure in pathological cases — especially where there is severe damage to the retinal outer segments. There is a need for a comprehensive volume of expert-developed training set data for diseased states, and validation of its accuracy in order for this method to be widely applicable in a clinical setting.
- Schmidt-Erfurth U, Leitgeb RA, Michels S, et al. Three-dimensional ultrahigh-resolution optical coherence tomography of macular diseases. Invest Ophthalmol Vis Sci. 2005;46(9):3393–3402. doi:10.1167/iovs.05-0370 [CrossRef]
- Sung KR, Sun JH, Na JH, Lee JY, Lee Y. Progression detection capability of macular thickness in advanced glaucomatous eyes. Ophthalmology. 2012;119(2):308–313. doi:10.1016/j.ophtha.2011.08.022 [CrossRef]
- Elman MJ, Bressler NM, Qin H, et al. Expanded 2-year follow-up of ranibizumab plus prompt or deferred laser or triamcinolone plus prompt laser for diabetic macular edema. Ophthalmology. 2011;118(4):609–614. doi:10.1016/j.ophtha.2010.12.033 [CrossRef]
- Schwartz SD, Hubschman JP, Heilwell G, et al. Embryonic stem cell trials for macular degeneration: A preliminary report. Lancet. 2012;379(9817):713–720. doi:10.1016/S0140-6736(12)60028-2 [CrossRef]
- Saidha S, Syc SB, Ibrahim MA, et al. Primary retinal pathology in multiple sclerosis as detected by optical coherence tomography. Brain. 2011;134(Pt 2):518–533. doi:10.1093/brain/awq346 [CrossRef]
- Mylonas G, Ahlers C, Malamos P, et al. Comparison of retinal thickness measurements and segmentation performance of four different spectral and time domain OCT devices in neovascular age-related macular degeneration. Br J Ophthalmol. 2009;93(11):1453–1460. doi:10.1136/bjo.2008.153643 [CrossRef]
- Kim SW, Oh J, Kwon SS, Yoo J, Huh K. Comparison of choroidal thickness among patients with healthy eyes, early age-related maculopathy, neovascular age-related macular degeneration, central serous chorioretinopathy, and polypoidal choroidal vasculopathy. Retina. 2011;31(9):1904–1911. doi:10.1097/IAE.0b013e31821801c5 [CrossRef]
- Saidha S, Sotirchos ES, Ibrahim MA, et al. Microcystic macular oedema, thickness of the inner nuclear layer of the retina, and disease characteristics in multiple sclerosis: A retrospective study. Lancet Neurol. 2012;11(11):963–972. doi:10.1016/S1474-4422(12)70213-2 [CrossRef]
- Mwanza JC, Oakley JD, Budenz DL, Chang RT, Knight OJ, Feuer WJ. Macular ganglion cell-inner plexiform layer: Automated detection and thickness reproducibility with spectral domain-optical coherence tomography in glaucoma. Invest Ophthalmol Vis Sci. 2011;52(11):8323–8329. doi:10.1167/iovs.11-7962 [CrossRef]
- Patel PJ, Foster PJ, Grossi CM, et al. Spectral-domain optical coherence tomography imaging in 67 321 adults: Associations with macular thickness in the UK Biobank Study. Ophthalmology. 2016;123(4):829–840. doi:10.1016/j.ophtha.2015.11.009 [CrossRef]
- Symes R, Young M, Forooghian F. Quantitative assessment of retinal degeneration in birdshot chorioretinopathy using optical coherence tomography. Ophthalmic Surg Lasers Imaging Retina. 2015;46(10):1009–1012. doi:10.3928/23258160-20151027-04 [CrossRef]
- Docherty G, Martens R, Forooghian F. Increased photoreceptor outer segment volume following systemic treatment of birdshot chorioretinopathy. Retin Cases Brief Rep. 2017Aug14. doi:. [Epub ahead of print] doi:10.1097/ICB.0000000000000622 [CrossRef]
- Bae SS, Forooghian F. Optical coherence tomography-based quantification of photoreceptor injury and recovery in Vogt-Koyanagi-Harada Uveitis. Ocul Immunol Inflamm. 2017;25(3):338–343. doi:10.3109/09273948.2015.1125510 [CrossRef]
- Forooghian F, Stetson PF, Meyer SA, et al. Relationship between photoreceptor outer segment length and visual acuity in diabetic macular edema. Retina. 2010;30(1):63–70. doi:10.1097/IAE.0b013e3181bd2c5a [CrossRef]
- Pappuru RR, Ouyang Y, Nittala MG, et al. Relationship between outer retinal thickness substructures and visual acuity in eyes with dry age-related macular degeneration. Invest Ophthalmol Vis Sci. 2011;52(9):6743–6748. doi:10.1167/iovs.10-6723 [CrossRef]
- Subhi Y, Forshaw T, Sørensen TL. Macular thickness and volume in the elderly: A systematic review. Ageing Res Rev. 2016;29:42–49. doi:10.1016/j.arr.2016.05.013 [CrossRef]
- Lang A, Carass A, Hauser M, et al. Retinal layer segmentation of macular OCT images using boundary classification. Biomed Opt Express. 2013;4(7):1133–1152. doi:10.1364/BOE.4.001133 [CrossRef]
- Tian J, Varga B, Tatrai E, et al. Performance evaluation of automated segmentation software on optical coherence tomography volume data. J Biophotonics. 2016;9(5):478–489. doi:10.1002/jbio.201500239 [CrossRef]
- Zhao MH, Wu Q, Hu P, Jia LL. Macular thickness in myopia: An OCT study of young Chinese patients. Curr Eye Res. 2016;41(10):1373–1378. doi:10.3109/02713683.2015.1119854 [CrossRef]
- Ooto S, Hangai M, Tomidokoro A, et al. Effects of age, sex, and axial length on the three-dimensional profile of normal macular layer structures. Invest Ophthalmol Vis Sci. 2011;52(12):8769–8779. doi:10.1167/iovs.11-8388 [CrossRef]
- Liu X, Shen M, Yuan Y, et al. Macular thickness profiles of intraretinal layers in myopia evaluated by ultrahigh-resolution optical coherence tomography. Am J Ophthalmol. 2015;160(1):53–61.e2. doi:10.1016/j.ajo.2015.03.012 [CrossRef]
- Chen M, Rajapakse D, Fraczek M, Luo C, Forrester JV, Xu H. Retinal pigment epithelial cell multinucleation in the aging eye — a mechanism to repair damage and maintain homoeostasis. Aging Cell. 2016;15(3):436–445. doi:10.1111/acel.12447 [CrossRef]
- Demirkaya N, van Dijk HW, van Schuppen SM, et al. Effect of age on individual retinal layer thickness in normal eyes as measured with spectral-domain optical coherence tomography. Invest Ophthalmol Vis Sci. 2013;54(7):4934–4940. doi:10.1167/iovs.13-11913 [CrossRef]
- Hardin JS, Taibbi G, Nelson SC, Chao D, Vizzeri G. Factors affecting Cirrus-HD OCT optic disc scan quality: A review with case examples. J Ophthalmol. 2015;2015:746150. doi:10.1155/2015/746150 [CrossRef]
General Linear Model Analysis for Photoreceptor Outer Segment Volume
|Parameters||Β||Standard Error||t-test Statistics||P Value|
|Gender (female vs. male)||0.04||0.022||1.804||.076|
|Vitreomacular adhesion(no vs. yes)||−0.026||0.028||−0.91||.367|