Open Access
Add to BibSonomy
Abstract
We assessed the ability of the optical attenuation coefficient (AC) to detect early-stage glaucoma with two AC estimation algorithms: retinal layer intensity ratio (LIR) and depth-resolved confocal (DRC). We also introduced new depth-dependent AC parameters for retinal nerve fiber layer assessment. Optical coherence tomography B-scans were collected from 44 eyes of age-similar participants with eye health ranging from healthy to severe glaucoma, including glaucoma suspect patients. Mean AC values estimated from the DRC method are comparable to ratio-extracted values (p > 0.5 for all study groups), and the depth-dependent ACDRC parameters enhance the utility of the AC for detection of early-stage glaucoma.
© 2022 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Glaucoma is the second leading cause of blindness globally [1]. Pathological changes in glaucomatous eyes include the progressive degeneration of retinal ganglion cells and the thinning of the retinal nerve fiber layer (RNFL), where the ganglion cell axons reside. This degenerative process can be detected clinically by measuring the thickness of the RNFL with optical coherence tomography (OCT), a critical technology for glaucoma diagnosis. OCT allows in vivo, non-invasive imaging of the retina with high resolution and diagnostic accuracy.
Although measurement of RNFL thickness effectively differentiates glaucomatous eyes from healthy eyes [2–4], a large variation exists among the thickness measurements in healthy individuals due to structural differences such as myopia [5–7] and optic disc size [8, 9]. This inherent overlap between the RNFL thicknesses of some healthy eyes and early-stage glaucomatous eyes leads to challenges in early glaucoma detection. As a result, researchers have sought to investigate whether other parameters, such as ganglion cell layer thickness, RNFL reflectance, and volumetric analysis of the retina, may be more reliable indicators of early-stage glaucoma [7, 10, 11].
In addition to the thinning of the RNFL that can be visualized on OCT B-scans, the death of retinal ganglion cell axons in glaucoma leads to changes in the scattering property of the RNFL [12, 13]. This change is hypothesized to be caused by the reduction in axon density in diseased RNFL [14–16], and it can be quantified through measurement of the optical attenuation coefficient (AC). The AC measures how quickly incident light is attenuated when passing through a medium and is a function of the underlying medium properties. AC values can serve as an indicator for microscopic changes within tissues [17–23].
Previous studies have shown a significant difference between ACRNFL values in healthy and glaucomatous eyes (p < 0.001) [13, 16]. However, the method used in those relies on determining the intensity ratio between two scattering layers – the RNFL and the retinal pigment epithelium (RPE). For convenience, the algorithm will be referred to as the layer intensity ratio (LIR) method. Unfortunately, LIR requires segmentation of the retina, an extra step in the processing pipeline that typically requires manual input. New variations of a depth-resolved method of extracting the AC have been introduced recently [24–26]. These methods remove the need for segmentation and increase the resolution of the AC measurement by allowing pixel-wise estimation [24], which can be important in detecting smaller changes that occur in early glaucoma. Our group has previously demonstrated a depth-resolved model that accounts for the confocal function (Depth-resolved confocal algorithm, or DRC) [25]. When combined with our algorithm for image-based extraction of the confocal function parameters (AutoConfocal) [27] the DRC allows for automated estimation of the AC without a priori knowledge of the imaging system parameters, which permits facile integration in the clinical workflow.
In this study, we explore the potential of the AutoConfocal + DRC method, henceforth referred to as the DRC method, to differentiate glaucomatous disease states in volunteers with normal vision, glaucoma suspects, and patients with mild and moderate/severe glaucoma. Our primary goal is to demonstrate the degree to which DRC can differentiate healthy from glaucomatous individuals compared to LIR algorithm. In addition, our inclusion of glaucoma suspect patients allows us to investigate the clinical value of the AC of the RNFL for early glaucoma detection, which has not been previously explored. Lastly, with the depth-resolved information we extracted from the DRC analysis, we introduce new depth-dependent parameters that show promise as potential diagnostic biomarkers.
2. Methods
2.1 Human subject recruitment
With approval from the Institutional Review Board of Vanderbilt University (IRB #190937), we recruited patients presenting to the Vanderbilt Eye Institute who were previously diagnosed with glaucoma or high risk of glaucoma based on standard-of-care methods, including a visual field test, biomicroscopy and OCT. Informed consent was obtained after thorough explanation of the study. Glaucoma severity was determined by the mean deviation (MD) of the visual field test (Humphrey field analyzer II, Carl Zeiss Meditec), where mild glaucoma cases had an MD ranging from 0 to -6.0 dB and moderate/severe cases had an MD worse than -6.0 dB. Eyes with visual acuity (VA) worse than 20/50 were excluded from analysis. Glaucoma suspects were identified by clinicians: these patients had high risk of glaucoma by showing some suspicious feature during exam, such as optic nerve asymmetry, a family history of glaucoma, or a high intraocular pressure (IOP), but had not shown functional damage on the visual field test or thinning of the RNFL on OCT scans. Age-similar healthy volunteers were recruited as a control population. For inclusion, healthy participants needed to have a healthy eye history (except for uncomplicated cataract surgery), IOP of 21 mm Hg or lower, no signs of glaucoma when examining the optic nerve appearance and 20/25 or better VA. Individuals with a visually significant cataract (without receiving surgery) were excluded from the study, to avoid poor signal quality.
2.2 Participant imaging
For each eye, we collected three B-scans in succession in a single sitting, with each of the three scans varying the depth of the focal plane to allow extraction of the confocal parameters from the AutoConfocal algorithm [27]. All scans were acquired on the Zeiss Cirrus 5000 Spectral Domain OCT machine (Carl Zeiss Meditec, Inc, Dublin, CA). A built-in eye-tracker was used during image acquisition to ensure that the scans were collected from the same location in the eye. The Cirrus scan was taken as one line of a 9-mm scan tilted at 7 degrees, centered on the optic nerve and the fovea.
2.3 Image processing and analysis
2.3.1 Manual segmentation
To demonstrate the comparative performance of DRC vs LIR for differentiating healthy from glaucomatous individuals, we performed manual segmentation of the inner and outer surface of the RNFL and of RPE of each B-scan, as required by the LIR algorithm. One annotator, while masked to clinical data, manually labeled the RNFL and RPE on a photo-editor software (Digital Ruby, LLC. YouDoodle Pro. Version 7.9.0) and then imported the labeled B-scans to a custom MATLAB (The Math Works, Inc. MATLAB. Version 2018b) algorithm for extraction of the layer information. The annotation was validated by another observer, who was also masked to the clinical information. Note that the outer RPE surface (purple line in Fig. 1, left column) was segmented by hand while the inner RPE surface (blue line) was marked by offsetting the outer surface by a fixed number of pixels, as has been done in prior work [16]. This method assumes that the RPE has a constant thickness in the B-scan. The extracted surfaces were smoothed with local regression using a second-degree polynomial model and a weighted linear least squares algorithm to remove non-biologically realistic features in the manual segmentation (e.g., sharp edges and straight-line segments).
2.3.2 Region selection
The target site of analysis was a 20-A-scan region located at a 1.3-mm distance from the center of the optic nerve head (ONH) [16], shown in the middle column of Fig. 1. This region was selected because it is the most clinically relevant region on the retina for RNFL assessment: most RNFL thickness measurements are performed on a ring-shaped region 1.3 mm or 1.7 mm away from the ONH [28,29].
2.3.3 AC calculations
The steps we took to obtain the AC estimations with the LIR method and the DRC method are illustrated in the right column of Fig. 1.
AC calculations with the LIR method: This method is a derivation based on the original curve-fitting AC algorithm introduced by Faber et al.: the light intensity is modeled as
where h is the confocal function, $h(z )= \; {\left( {{{\left( {\frac{{z - {z_{cf}}}}{{{z_R}}}} \right)}^2} + 1} \right)^{ - 1}}$, and zcf and zR denotes the system parameters, the focal plane depth and the apparent Rayleigh range, respectively. In the papers by Vermeer [13, 30] and van der Schoot [16] in 2012, they extended the original curve-fitting method to extract the AC from the RNFL, utilizing the RPE as the reference layer, where intensity ratios of the RNFL and RPE can be expressed as:In total we calculated 20 AC measurements from the 20-A-scan region for every B-scan. We then determined one AC value for each B-scan by averaging the 20 A-scan AC measurements. Lastly, we averaged the AC measurements from the three B-scans we collected from each eye and used the averaged AC measurement for statistical analysis.
AC calculations with the DRC method: We generated AC mappings for the corresponding B-scans with the DRC algorithm. The DRC is based on the original depth-resolved method, introduced by Vermeer et al. in 2014. Here, the light intensity is modeled as
where A is the backscattering ratio. The AC, accounting for a maximum imaging depth D, can be solved asThe discretized AC measurement for each pixel i with pixel size $\mathrm{\Delta }$ is given by
For a direct comparison with the LIR results, the same 20-pixel-wide region was selected from the AC mapping for estimation. Because the DRC method provided pixel-wise calculation of AC (that is, it produces an AC value for every pixel within an A-scan), we took the mean of the depth-dependent ACs in the RNFL region for each A-scan to yield 20 averaged A-scan AC measurements per B-scan. Then, we calculated one AC value for each B-scan by averaging the 20 A-scan AC measurements. Lastly, we determined an averaged AC measurement for each eye from its three B-scans.
2.4 Statistical analysis
Statistical analysis of differences between pairs of study groups on the testing parameters (ACLIR, ACDRC and depth-dependent AC parameters) were based on the nonparametric Mann-Whitney U test. Comparisons between the LIR and DRC methods were done through the Wilcoxon matched pair test. To compare the patient information and scan quality across all study groups, we used Kruskal-Wallis test, where Gaussian distributions were not assumed. VA measurements were converted into LogMAR scale for statistical analysis.
3. Results
3.1 Participant demographics
We imaged and analyzed 44 eyes from healthy volunteers and from patients presenting to the Vanderbilt Eye Institute who were previously diagnosed with glaucoma or high risk of glaucoma. Of the 44 eyes, 10 were considered glaucoma suspects, 12 had mild-stage glaucoma, 12 had moderate- or severe-stage glaucoma, and 10 were healthy controls. Note that four healthy eyes analyzed in the study have had cataract surgery and had an intraocular lens implant present at the time of imaging. Table 1 shows the mean and standard deviation of patient age, IOP and VA of eyes included in each study group. The mean ages of the healthy group, glaucoma suspect group, mild glaucoma group, and moderate/severe glaucoma group were 72.3, 71.5, 70.4, and 71.2, respectively. There were no significant differences among the mean ages of the study groups (p = 0.85) and thus we considered our study groups as age-similar. Therefore, although aging is known to contribute to changes in the RNFL [31], we can assume that the changes in AC that we observed among different study groups were not related to age but, rather, to disease progression. The IOP did not differ significantly across the study groups. In contrast, VA showed correlation with disease severity, as patients may have worse VA at more advanced glaucoma stages. We also recorded the signal strength of the scans, and the lowest recorded signal strength was 6/10. We did observe signal strength variations across the study groups. However, we did not expect that the differences in the signal strength would affect our measurements, as previous studies have shown that the small changes observed in the RNFL thickness when the signal strengths drop from 10/10 to 6/10 are not considered clinically significant [32].
Table 1. Study subject statistics: Mean and standard deviation of age, intraocular pressure (IOP), visual acuity (VA), signal strength and sample size of each study group. The p-values, calculated for each row, indicate whether there are statistically significant differences among the study groups (p ≤ 0.05 suggests statistically significant differences).
3.2 DRC vs. LIR
We studied whether the DRC method had comparable performance and produced similar results to LIR. Representative OCT B-scans and the corresponding AC mappings from healthy, glaucoma suspect, mild glaucoma and moderate/severe glaucoma eyes are shown in the top two rows of Fig. 2. When comparing across the disease stages, the AC mapping allows direct visualization of disease progression, with decreasing AC values in the more severely impacted RNFL. The bottom row of Fig. 2 shows the depth-dependent AC measurements (solid pink line) and mean AC (dashed pink line) from the DRC method and the single-valued AC calculated from the LIR method (solid blue line) for the A-scan denoted with an arrow in the middle row.
Fig. 2. OCT B-scans and corresponding AC mappings of healthy (OS), glaucoma suspect (OS), mild glaucoma (OS) and moderate/severe glaucoma eyes (OS); bottom row shows extracted depth-dependent AC value by the DRC method and LIR method from the A-scan indicated by the white arrow. OS: left eye. Scale bar: 0.25 mm.
DRC and LIR yield comparable AC estimations without statistically significant differences (p > 0.5 for all severity groups). The average AC measurements, ACDRC and ACLIR, are plotted in Fig. 3 for the DRC method and LIR method, respectively. The estimated AC values from ACDRC and ACLIR are 4.15 (95% C.I. 3.72 - 4.58) vs. 3.91 (95% C.I. 3.39 - 4.43) mm-1, 2.98 (95% C.I. 2.25–3.71) vs. 2.84 (95% C.I. 2.05–3.62) mm-1, 3.13 (95% C.I. 2.74–3.52) vs. 2.84 (95% C.I. 2.28–3.40) mm-1 and 2.37 (95% C.I. 1.82–2.91) vs. 2.16 (95% C.I. 1.58–2.74) mm-1 for healthy, glaucoma suspect, mild glaucoma and moderate/severe stage glaucoma, respectively. Note that the ACs for both methods are able to distinguish healthy eyes from glaucoma suspects and glaucomatous eyes.
Fig. 3. ACs from DRC and LIR and the calculated differences (absolute value shown in plot) between the two methods in healthy and glaucomatous eyes. * p < 0.05, *** p < 0.005, ns: no significance. p values shown on plot are calculated based on ACDRC. Error bars represent standard deviations.
Table 2 reports the estimated AC values from the two algorithms for each severity group and tests for statistical differences between the results from the two algorithms. Differences between the groups were not statistically significant (p value < 0.05 is considered significantly different).
Table 2. Mean RNFL attenuation coefficient calculated from layer intensity ratio (ACLIR) and depth-resolved (ACDRC) methods measured in healthy eyes and glaucomatous eyes [mm-1]a
3.3 Depth-dependent AC
A main advantage of the DRC method is that the AC can be calculated per pixel, which provides depth-dependent information. By plotting pixel-wise AC measurements, we can visually observe changes in the AC at different depths in the RNFL, as shown in the bottom row of Fig. 2. For most regions, we noticed a roughly linear, increasing trend in the AC vs. depth relationship, with some amplitude fluctuation (larger fluctuations in healthier RNFLs and smaller fluctuations in the RNFLs of more advanced glaucoma severity). To quantify the rate of increase and the amount of fluctuation of the AC measurements along depth, we applied linear regression to define two parameters to describe the local changes of the AC: the slope and the fitting residual.
For a given A-scan, we first normalized the depth vector, whose range comprises the thickness of the RNFL. Next, we performed linear regression on the depth-resolved AC as a function of normalized depth for each A-scan, where we determined a fitting slope and an averaged absolute fitting residual from each. An example plot of the depth-resolved AC from a single A-scan and the linear fit is shown in Fig. 4 (a). To reduce the fitting error and find the most linear region, five pixels near the top and bottom surface of the RNFL were excluded from the analysis, as AC values tend to decrease dramatically near the layer interface. As a result, two eyes from the moderate/severe group did not have sufficient data points in the RNFL for performing the linear regression and therefore were excluded from analysis. Then, we repeated the same calculation for all 20 A-scans in the region of interest and computed a mean fitting slope and absolute fitting residual for each B-scan. Lastly, we determined the values of depth-dependent parameters for a given eye by averaging the measurements from its three B-scans and used the averaged values in the statistical analysis.
Fig. 4. Depth-dependent AC analysis of the RNFL. (a) Depth-resolved AC plotted against normalized depth. (b) Averaged fitting slope of each study group. (c) Averaged absolute fitting residual of each study group. * p < 0.05, ** p < 0.01, *** p < 0.005.
Our results indicate that there is more rapid increase and larger fluctuations in the pixel-wise ACs as a function of depth in the healthy RNFL than the diseased RNFL. The mean fitting slope and fitting residual for each study group are shown in Fig. 4 (b) and (c). The fitting slope measurement can differentiate suspect eyes and mild glaucoma eyes from the healthy eyes (p < 0.01 and p < 0.005, respectively) with higher statistical significances than the mean AC measurements (p < 0.05). The statistical significance improves when differentiating the suspect eyes from the healthy with the fitting residual (p < 0.005).
4. Discussion
In this study, we applied depth-resolved AC analysis with the DRC algorithm to the detection of early-stage glaucoma and showed that it has comparable performance to the LIR method when analyzed with averaged AC values. The results in Fig. 3 and Table 2 do not show statistically significant differences when comparing the ACs extracted by DRC and LIR method from each study group, which suggests that DRC is as effective at predicting ACs from the RNFL as LIR and can be used to derived diagnostic metrics for evaluating pathological changes.
A previous study has shown that the AC can successfully differentiate glaucomatous eyes from healthy ones [13]. However, the disease population in that study was restricted to moderate glaucoma. In our study, we show that the mean AC is a valuable metric for detecting even early-stage glaucoma by including a mild glaucoma group as well as a glaucoma suspect group. Comparing the mean AC across study groups, we observe a decrease in the measurement as the disease stage advances. This decrease can be explained by the reduced density of nerve fibers and retinal capillaries, both of which have been shown in the progression of glaucoma [13,33]. This explanation is also supported by the finding of the decrease in the RNFL birefringence in glaucoma patients, as the alignment of axons is disrupted when axonal death occurs [10,34].
The ability of the DRC to extract depth-wise information enables us to extract new metrics (the slope and the residual) to describe the changes of AC along depth. We find that the information extracted from the depth-dependent AC may add diagnostic value. We reason that the depth-dependent increase in AC near the outer surface of the RNFL is due to the presence of radial peripapillary capillaries (RPC), which are long, straight vessels that run parallel with the RNFL axons and are located near the RNFL and ganglion cell layer interface [35, 36]. In the study of Cambell et al., the authors reported an increasing capillary density as a function of depth in the RNFL, which agrees with the trend we observed in the depth-resolved AC data [35]. Therefore, we reason that the AC in the RNFL is associated with a combination of scattering and absorption of light, where the RNFL axons contribute to the scattering and the blood in the RPC contributes to the absorption. The decrease in the slope of the depth-resolved AC in glaucoma takes into the account the reduction in the densities of both the nerve axon as well as the capillaries. The decrease in the fluctuations of the AC along depth in more severe glaucoma cases may be explained by the loss of RPC density due to glaucoma progression, as the capillary density has been reported as significantly lower in glaucomatous eyes than healthy eyes [33, 37, 38]. It is important to note that both the original depth-resolved algorithm and DRC make the assumption that the light intensity decays to zero at the end of the imaging range [24,25]. However, when such assumption is not true, the resulting AC values suffer from over-estimation for pixels near the bottom of the image [26,39]. In this paper, the retinal structure of interest, the RNFL, was placed in the top half of each B-scan; therefore, our analysis only minimally suffers by not implementing an end-of-range correction. Note that this issue would be problematic if one were to study tissue structures located near the end of the imaging range, as the AC values could peak and lead to erroneous mean AC, as well as an erroneous depth-dependent AC analysis. Solutions have been proposed to address this issue, where over-estimation is corrected for pixels near the bottom of the image [26]. This correction should be incorporated into depth-resolved AC extraction algorithm in future studies for more accurate estimations.
Many new diagnostics biomarkers with potential to detect earlier changes in the RNFL than the thickness measurement have been proposed in recent literature, including RNFL birefringence [10,34,40] and reflectance [14,41], which are tissue-optical-property-based metrics. Although studies have shown promising results of the RNFL birefringence in differentiation of early-stage glaucoma [34,40], this method relies on the use of polarization-sensitive OCT, which greatly hinders its potential for clinical translation. The AC, which is also based on tissue optical properties, may be a complementary or even superior metric to the clinical standards, such as the RNFL thickness measurement. A previous study has demonstrated the structure-function correlation of AC to the mean deviation of the visual field test in glaucoma population [16]. When compared to thickness-based analysis, such as in the RNFL or GCL, the AC measurement may be less affected by natural biological variation [6,42], as the ACs are based on the contrast of internal tissue structures. When compared to other emerging tissue-optical-property metrics, the DRC method we use here can be performed on currently available OCT systems that are already widely distributed in clinical practices worldwide.
A major advantage of the DRC algorithm is that it does not require segmentation of the retinal layers (i.e. RNFL, RPE, or GCL) to obtain attenuation coefficient measurements: the spatially resolved information of potential diagnostic value. Moreover, it is robust against cases of a thin RNFL or irregularly shaped retina; whereas, segmentation of the RNFL is more likely to fail and would result in inaccurate calculation of the attenuation coefficients using the curve-fitting or the LIR algorithm. While automated segmentation could remove a barrier to LIR implementation, DRC still holds an advantage that it enables visualization of pixel-wise changes of the RNFL in cross-section, which may enable detection of earlier glaucomatous changes that would otherwise be missed from a regular OCT image or an en face AC mapping. In cases where an averaged regional reading from the AC map, or quantitative, depth-dependent AC measurements were to be desired, AC-enabled auto-segmentation could be easily incorporated, and has been shown to yield superior performance for segmentation than its intensity-based alternatives [43,44].
Additionally, when compared to the RNFL reflectance, the depth-dependent AC analysis may be a more reliable measurement because it measures the rate of signal decay and is independent of the signal strength [12]. Both the reflectivity and the birefringence of the RNFL have been shown to be more sensitive to disease progression than thickness measurements in glaucoma, and it was reasoned in the literature that the disruption of the integrity of the axonal cytoskeleton caused by glaucoma could lead to changes in the reflectivity and birefringence in early onset glaucoma and even glaucoma suspect patients, which precedes noticeable reduction in the RNFL thickness [10,14]. Since the AC is the measurement of the rate of light attenuation and is inherently correlated with reflectivity and tissue birefringence, we expect that the AC, too, could serve as a diagnostic parameter to enable earlier detection and more accurate monitoring of glaucoma.
Lastly, our study reports the AC from glaucoma suspect patients, which is a population that was omitted by previous studies on the AC. The AC measured from this group falls between the ACs of the normal and the confirmed glaucoma population. Longitudinal studies on glaucoma suspect patients may reveal the potential of AC and depth-resolved AC analysis for establishing a threshold for the possibility of the actual development of glaucoma in these patients. The health care benefits of deploying this analysis are that it could serve as a form of triage that could permit more efficient dedication of clinical resources to follow-up and provide timely treatment to true glaucoma patients, as the current clinic is overpopulated by glaucoma suspects.
A limitation of this study is that, due to the availability of data, we only acquired high-resolution images and corresponding analysis of a limited area of the RNFL: the temporal region. This limitation may explain the overlap of AC values between the healthy and early-stage glaucoma eyes, as glaucoma may have affected the superior or the inferior quadrants more dramatically than the area where we took the measurement. However, the promising results shown in the depth-dependent AC analysis encourage future studies in all quadrants of the optic nerve with a greater number of patients. Another limitation was the use of manual segmentation to perform the analysis. This was done because reliable automatic segmentation of the RPE for the high-resolution line scan data was not available at the time of the experiment. However, such algorithms are widely available for volumetric data for the ONH. As we proceed to the collection of volumetric data for further research, we will adopt automatic segmentation approaches for more efficient analysis.
In summary, the use of the AC as extracted from the DRC has strong potential to assist with clinical evaluation of early-stage glaucoma. To this end, a strong advantage of the method we proposed in the paper is that it introduces minimal changes to the current clinical workflow, as it does not require extra hardware or extensive data collection. A volumetric analysis of depth-dependent AC in the entire region of the ONH is beyond the scope of this study. In future studies, we will acquire retinal volume scans from a larger number of participants, extract the ACRNFL from all quadrants for analysis and build a classifier for assessing the performance of these newly defined metrics against the traditional ones, such as thickness measurements.
Acknowledgments
The authors acknowledge help from Dr. Eric Brown on statistical analysis. The authors thank Dr. Luis de Sisternes Garcia for helpful conversations and reviews of an early version of this manuscript.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request
References
1. J. D. Adelson, R. R. A. Bourne, P. S. Briant, et al., “Causes of blindness and vision impairment in 2020 and trends over 30 years, and prevalence of avoidable blindness in relation to VISION 2020: the Right to Sight: an analysis for the Global Burden of Disease Study,” Lancet Glob. Heal. 9(2), e144–e160 (2020).
2. C. K. Leung, C. Y. Cheung, R. N. Weinreb, Q. Qiu, S. Liu, H. Li, G. Xu, N. Fan, L. Huang, C.-P. Pang, and D. S. C. Lam, “Retinal Nerve Fiber Layer Imaging with Spectral-Domain Optical Coherence Tomography: A Variability and Diagnostic Performance Study,” Ophthalmology 116(7), 1257–1263.e2 (2009). [CrossRef]
3. A. Kanamori, M. Nakamura, M. F. T. Escano, R. Seya, H. Maeda, and A. Negi, “Evaluation of the glaucomatous damage on retinal nerve fiber layer thickness measured by optical coherence tomography,” Am. J. Ophthalmol. 135(4), 513–520 (2003). [CrossRef]
4. F. A. Medeiros, L. M. Zangwill, C. Bowd, R. M. Vessani, R. Susanna, and R. N. Weinreb, “Evaluation of retinal nerve fiber layer, optic nerve head, and macular thickness measurements for glaucoma detection using optical coherence tomography,” Am. J. Ophthalmol. 139(1), 44–55 (2005). [CrossRef]
5. M. J. Kim, E. J. Lee, and T. W. Kim, “Peripapillary retinal nerve fibre layer thickness profile in subjects with myopia measured using the Stratus optical coherence tomography,” Br. J. Ophthalmol. 94(1), 115–120 (2010). [CrossRef]
6. S. H. Kang, S. W. Hong, S. K. Im, S. H. Lee, and M. D. Ahn, “Effect of myopia on the thickness of the retinal nerve fiber layer measured by cirrus HD optical coherence tomography,” Invest. Ophthalmol. Visual Sci. 51(8), 4075–4083 (2010). [CrossRef]
7. S. Seo, C. E. Lee, J. H. Jeong, K. H. Park, D. M. Kim, and J. W. Jeoung, “Ganglion cell-inner plexiform layer and retinal nerve fiber layer thickness according to myopia and optic disc area: A quantitative and three-dimensional analysis,” BMC Ophthalmol. 17(1), 1–8 (2017). [CrossRef]
8. G. Savini, M. Zanini, V. Carelli, A. A. Sadun, F. N. Ross-Cisneros, and P. Barboni, “Correlation between retinal nerve fibre layer thickness and optic nerve head size: An optical coherence tomography study,” Br. J. Ophthalmol. 89(4), 489–492 (2005). [CrossRef]
9. D. L. Budenz, D. R. Anderson, R. Varma, J. Schuman, L. Cantor, J. Savell, D. S. Greenfield, V. M. Patella, H. A. Quigley, and J. Tielsch, “Determinants of normal retinal nerve fiber layer thickness measured by Stratus OCT,” Ophthalmology 114(6), 1046–1052 (2007). [CrossRef]
10. X.-R. Huang, R. W. Knighton, Y. Z. Spector, J. Qiao, W. Kong, and Q. Zhao, “Reflectance spectrum and birefringence of the retinal nerve fiber layer with hypertensive damage of axonal cytoskeleton,” Invest. Ophthalmol. Visual Sci. 58(4), 2118–2129 (2017). [CrossRef]
11. J. Xu, H. Ishikawa, G. Wollstein, R. A. Bilonick, L. S. Folio, Z. Nadler, L. Kagemann, and J. S. Schuman, “Three-dimensional spectral-domain optical coherence tomography data analysis for glaucoma detection,” PLoS One 8(2), e55476 (2013). [CrossRef]
12. S. K. Gardiner, S. Demirel, J. Reynaud, and B. Fortune, “Changes in retinal nerve fiber layer reflectance intensity as a predictor of functional progression in Glaucoma,” Invest. Ophthalmol. Visual Sci. 57(3), 1221–1227 (2016). [CrossRef]
13. K. A. Vermeer, J. van der Schoot, H. G. Lemij, and J. F. de Boer, “RPE-normalized RNFL attenuation coefficient maps derived from volumetric OCT imaging for glaucoma assessment,” Invest. Ophthalmol. Visual Sci. 53(10), 6102–6108 (2012). [CrossRef]
14. X.-R. Huang, Y. Zhou, W. Kong, and R. W. Knighton, “Reflectance decreases before thickness changes in the retinal nerve fiber layer in glaucomatous retinas,” Invest. Ophthalmol. Visual Sci. 52(9), 6737–6742 (2011). [CrossRef]
15. M. E. Pons, H. Ishikawa, R. Gürses-Özden, J. M. Liebmann, H. L. Dou, and R. Ritch, “Assessment of retinal nerve fiber layer internal reflectivity in eyes with and without glaucoma using optical coherence tomography,” Arch. Ophthalmol. 118(8), 1044 (2000). [CrossRef]
16. J. van der Schoot, K. A. Vermeer, J. F. de Boer, and H. G. Lemij, “The effect of glaucoma on the optical attenuation coefficient of the retinal nerve fiber layer in Spectral Domain Optical Coherence Tomography images,” Invest. Ophthalmol. Visual Sci. 53(4), 2424–2430 (2012). [CrossRef]
17. B. G. Muller, R. A. A. van Kollenburg, A. Swaan, E. C. H. Zwartkruis, M. J. Brandt, L. S. Wilk, M. Almasian, A. W. Schreurs, D. J. Faber, L. R. Rozendaal, A. N. Vis, J. A. Nieuwenhuijzen, J. R. J. A. van Moorselaar, J. J. M. C. H. de la Rosette, D. M. de Bruin, and T. G. van Leeuwen, “Needle-based optical coherence tomography for the detection of prostate cancer: a visual and quantitative analysis in 20 patients,” J. Biomed. Opt. 23(08), 1 (2018). [CrossRef]
18. J. E. Freund, D. J. Faber, M. T. Bus, T. G. van Leeuwen, and D. M. de Bruin, “Grading upper tract urothelial carcinoma with the attenuation coefficient of in-vivo optical coherence tomography,” Lasers Surg. Med. 51, 399–406 (2019). [CrossRef]
19. S. Liu, “Tissue characterization with depth-resolved attenuation coefficient and backscatter term in intravascular optical coherence tomography images,” J. Biomed. Opt. 22(09), 1 (2017). [CrossRef]
20. C. Xu, J. M. Schmitt, S. G. Carlier, and R. Virmani, “Characterization of atherosclerosis plaques by measuring both backscattering and attenuation coefficients in optical coherence tomography,” J. Biomed. Opt. 13(3), 034003 (2008). [CrossRef]
21. Q. Zhao, C. Zhou, H. Wei, Y. He, X. Chai, and Q. Ren, “Ex vivo determination of glucose permeability and optical attenuation coefficient in normal and adenomatous human colon tissues using spectral domain optical coherence tomography,” J. Biomed. Opt. 17(10), 1050041 (2012). [CrossRef]
22. U. Baran, Y. Li, R. K. Wang, U. T. K. U. B. Aran, and Y. L. I. Uandong, “In vivo tissue injury mapping using optical coherence tomography based methods,” Appl. Opt. 54(21), 6448–6453 (2015). [CrossRef]
23. C. L. R. Rodriguez, J. I. Szu, M. M. Eberle, Y. Wang, M. S. Hsu, D. K. Binder, and B. H. Park, “Decreased light attenuation in cerebral cortex during cerebral edema detected using optical coherence tomography,” Neurophotonics 1(2), 025004 (2014). [CrossRef]
24. K. A. Vermeer, J. Mo, J. J. A. Weda, H. G. Lemij, and J. F. de Boer, “Depth-resolved model-based reconstruction of attenuation coefficients in optical coherence tomography,” Biomed. Opt. Express 5(1), 322 (2014). [CrossRef]
25. G. T. Smith, N. Dwork, D. O’Conner, U. Sikora, K. L. Lurie, J. M. Pauly, and A. K. Ellerbee, “Automated, Depth Resolved Estimation of Attenuation Coefficient From Optical Coherence Tomography Data,” IEEE Trans. Med. Imaging 34(12), 2592–2602 (2015). [CrossRef]
26. J. Liu, N. Ding, Y. Yu, X. Yuan, S. Luo, J. Luan, Y. Zhao, Y. Wang, and Z. Ma, “Optimized depth-resolved estimation to measure optical attenuation coefficients from optical coherence tomography and its application in cerebral damage determination,” J. Biomed. Opt. 24(03), 1–11 (2019). [CrossRef]
27. N. Dwork, G. T. Smith, T. Leng, J. M. Pauly, and A. K. Bowden, “Automatically determining the confocal parameters from OCT B-Scans for quantification of the attenuation coefficients,” IEEE Trans. Med. Imaging 38(1), 261–268 (2019). [CrossRef]
28. A. L. Rothman, M. B. Sevilla, S. F. Freedman, A. Y. Tong, V. Tai, D. Tran-Viet, S. Farsiu, C. A. Toth, and M. A. El-Dairi, “Assessment of retinal nerve fiber layer thickness in healthy, full-term neonates,” Am. J. Ophthalmol. 159(4), 803–811.e2 (2015). [CrossRef]
29. S. Zotter, M. Pircher, E. Götzinger, T. Torzicky, H. Yoshida, F. Hirose, S. Holzer, J. Kroisamer, C. Vass, U. Schmidt-Erfurth, and C. K. Hitzenberger, “Measuring retinal nerve fiber layer birefringence, retardation, and thickness using wide-field, high-speed polarization sensitive spectral domain OCT,” Invest. Ophthalmol. Visual Sci. 54(1), 72–84 (2013). [CrossRef]
30. K. A. Vermeer, J. van der Schoot, H. G. Lemij, and J. F. de Boer, “Quantitative RNFL attenuation coefficient measurements by RPE-normalized OCT data,” Ophthalmic Technol. XXII 8209, 82090U (2012). [CrossRef]
31. R. S. Parikh, S. R. Parikh, G. C. Sekhar, S. Prabakaran, J. G. Babu, and R. Thomas, “Normal Age-Related Decay of Retinal Nerve Fiber Layer Thickness,” Ophthalmology 114(5), 921–926 (2007). [CrossRef]
32. C. Samarawickrama, A. Pai, S. C. Huynh, G. Burlutsky, T. Y. Wong, and P. Mitchell, “Influence of OCT signal strength on macular, optic nerve head, and retinal nerve fiber layer parameters,” Invest. Ophthalmol. Visual Sci. 51(9), 4471–4475 (2010). [CrossRef]
33. T. Mansoori, J. Gamalapati, J. Sivaswamy, and N. Balakrishna, “Optical coherence tomography angiography measured capillary density in the normal and glaucoma eyes,” Saudi J. Ophthalmol. 32(4), 295–302 (2018). [CrossRef]
34. S. Desissaire, A. Pollreisz, A. Sedova, D. Hajdu, F. Datlinger, S. Steiner, C. Vass, F. Schwarzhans, G. Fischer, M. Pircher, U. Schmidt-Erfurth, and C. K. Hitzenberger, “Analysis of retinal nerve fiber layer birefringence in patients with glaucoma and diabetic retinopathy by polarization sensitive OCT,” Biomed. Opt. Express 11(10), 5488 (2020). [CrossRef]
35. J. P. Campbell, M. Zhang, T. S. Hwang, S. T. Bailey, D. J. Wilson, Y. Jia, and D. Huang, “Detailed Vascular Anatomy of the Human Retina by Projection-Resolved Optical Coherence Tomography Angiography,” Sci. Rep. 7(1), 1–11 (2017). [CrossRef]
36. T. Mase, A. Ishibazawa, T. Nagaoka, H. Yokota, and A. Yoshida, “Radial peripapillary capillary network visualized using wide-field montage optical coherence tomography angiography,” Invest. Ophthalmol. Visual Sci. 57(9), OCT504 (2016). [CrossRef]
37. C. L. Chen, A. Zhang, K. D. Bojikian, J. C. Wen, Q. Zhang, C. Xin, R. C. Mudumbai, M. A. Johnstone, P. P. Chen, and R. K. Wang, “Peripapillary retinal nerve fiber layer vascular microcirculation in glaucoma using optical coherence tomography–based microangiography,” Invest. Ophthalmol. Visual Sci. 57(9), OCT475 (2016). [CrossRef]
38. A. Yarmohammadi, L. M. Zangwill, A. Diniz-Filho, M. H. Suh, P. I. Manalastas, N. Fatehee, S. Yousefi, A. Belghith, L. J. Saunders, F. A. Medeiros, D. Huang, and R. N. Weinreb, “Optical coherence tomography angiography vessel density in healthy, glaucoma suspect, and glaucoma eyes,” Invest. Ophthalmol. Visual Sci. 57(9), OCT451 (2016). [CrossRef]
39. S. Chang and A. K. Bowden, “Review of methods and applications of attenuation coefficient measurements with optical coherence tomography,” J. Biomed. Opt. 24(09), 1 (2019). [CrossRef]
40. E. Gotzinger, M. Pircher, B. Baumann, H. Resch, C. Vass, and C. K. Hitzenberger, “Comparison of Retinal Nerve Fiber Layer Birefringence and Thickness of Healthy and Glaucoma Suspect Eyes Measured with Polarization Sensitive Spectral Domain OCT,” Invest. Ophthalmol. Vis. Sci. 50(13), 5823 (2009).
41. X.-R. Huang, R. W. Knighton, W. J. Feuer, and J. Qiao, “Retinal nerve fiber layer reflectometry must consider directional reflectance,” Biomed. Opt. Express 7(1), 22 (2016). [CrossRef]
42. J. K. Chung and Y. C. Yoo, “Correct calculation circle location of optical coherence tomography in measuring retinal nerve fiber layer thickness in eyes with myopic tilted discs,” Invest. Ophthalmol. Visual Sci. 52(11), 7894–7900 (2011). [CrossRef]
43. J. Novosel, Z. Wang, H. De Jong, M. Van Velthoven, K. A. Vermeer, and L. J. Van Vliet, “Locally-adaptive loosely-coupled level sets for retinal layer and fluid segmentation in subjects with central serous retinopathy,” Proc. - Int. Symp. Biomed. Imaging 2016(April), 702–705 (2016). [CrossRef]
44. T. Callewaert, J. Dik, and J. Kalkman, “Segmentation of thin corrugated layers in high-resolution OCT images,” Opt. Express 25(26), 32816 (2017). [CrossRef]
