Structure-function relationship in glaucoma: Optical coherence tomography en face imaging vs. red-free fundus photography

To analyse retinal nerve fibre layer (RNFL) defect measurements obtained from red-free fundus photography and optical coherence tomography (OCT) en face imaging, respectively, and to compare them for the strength of the structure–function association. Two hundred and fifty-six glaucomatous eyes of 256 patients with localized RNFL defect on red-free fundus photography were enrolled. A subgroup analysis included 81 highly myopic eyes (≤ –6.0 dioptres). Angular width of RNFL defect was compared between red-free fundus photography (i.e., red-free RNFL defect) and OCT en face imaging (i.e., en face RNFL defect). The correlation between angular width of each RNFL defect and functional outcomes, reported as mean deviation (MD) and pattern standard deviation (PSD), were assessed and compared. The angular width of en face RNFL defect was measured smaller than that of red-free RNFL defect in 91.0% eyes (mean difference, 19.98°). The association of en face RNFL defect with MD and PSD was stronger (R2 = 0.311 and R2 = 0.372, respectively) than that of red-free RNFL defect with MD and PSD (R2 = 0.162 and R2 = 0.137, respectively) (P < 0.05 for all). Especially in highly myopic eyes, the association of en face RNFL defect with MD and PSD was much stronger (R2 = 0.503 and R2 = 0.555, respectively) than that of red-free RNFL defect with MD and PSD (R2 = 0.216 and R2 = 0.166, respectively) (P < 0.05 for all). En face RNFL defect showed a higher correlation with severity of visual field loss than did red-free RNFL defect. The same dynamic was observed for highly myopic eyes.


INTRODUCTION
Evaluation of the retinal nerve fibre layer (RNFL) is crucial to detection of glaucoma [1].Red-free fundus photography has been used as the standard approach for detection of RNFL defect (i.e.red-free RNFL defect) [2].However RNFL loss often is not detectable until 50% of thickness has been lost [3].Thus, obtaining RNFL photographs of sufficient quality for interpretation is not infrequently difficult [1,4].For example, red-free RNFL defect can be difficult to be visualized in eyes with light pigment epithelial background and high myopia [5].Also, it might be obscured due to media opacity or small pupils.Another hurdle is that red-free RNFL defects do not always correspond to visual field (VF) defects for clear indications of structure-function relationship [6].All things considered, it would be of great clinical advantage to have a more sophisticated imaging tool that could predict functional damage more accurately.
Improved in vivo imaging has made it possible to identify the details of RNFL bundle projections.En face images from optical coherence tomography (OCT) scans using OCT angiography (OCTA) can display images of RNFL bundles for a full range of depths [7].Recent studies with OCT en face imaging have found localized RNFL defects in glaucomatous eyes (i.e.en face RNFL defects) [8][9][10][11][12].Hood et al. reported that missing or overlooked details of glaucomatous damage on traditional OCT RNFL thickness analysis were seen on OCT en face images [8].Moreover, other studies have presented good topographic correlations of en face RNFL defects with red-free RNFL defects [9][10][11].On the other hand, in eyes with diffuse RNFL defects, the details of both images showed poor topographic correlation [12].Diffuse red-free RNFL defects presented as a localized defect or no defect on OCT en face images.Under this circumstance, OCT en face imaging may yield clues to the detection of glaucomatous damage in difficult visualization of red-free fundus photography, such as high myopia.Also, RNFL defects on red-free fundus photography and OCT en face imaging may show discrepancies and/or divergent strengths of structure-function association.
Accordingly, objectives of the present study were to investigate the correlation and agreement of RNFL defects obtained with redfree fundus photography and OCT en face imaging and to compare the respective strengths of the structure-function association in glaucomatous eyes including highly myopic eyes.

Study subjects
The present study was based on the Gangnam Eye Cohort Study, conducted by Seoul National University Hospital (SNUH) Healthcare System Gangnam Center (HSGC).Detailed information on this cohort has been published elsewhere [13].This study was approved by the Seoul National University Hospital Institutional Review Board (IRB No. H-1906-141-1043).All of the procedures adhered to the tenets of the Declaration of Helsinki.Written informed consent to participate was obtained from all of the participants.
The study population comprises Korean subjects who had participated in a glaucoma screening program at SNUH HSGC and were enrolled when diagnosed with glaucoma at the Glaucoma Outpatient Clinic between January 2019 and November 2021.
Subjects with clear characteristic glaucomatous structural anomalies were included: [14] presence of glaucomatous optic disc change (e.g., increased cup-to-disc ratio, neuroretinal rim thinning or notching); single localized RNFL defect on red-free fundus photography; open angle; with or without corresponding VF loss.Glaucomatous VF defect was defined as (1) glaucoma hemifield test values outside the normal limits or (2) three or more abnormal contiguous points with a probability of P < 0.05, of which at least one point has a probability of P < 0.01 on a pattern deviation plot, or (3) a pattern standard deviation of P < 0.05.
Eligible subjects exhibited age ≥ 18 years, best-corrected visual acuities of 20/40 or better, mean deviation ≥ −12 dB in standard automated perimetry (SAP).Subjects were excluded with history of intraocular surgery aside from uncomplicated cataract surgery; secondary causes of glaucoma; coexisting retinal disease, uveitis, or nonglaucomatous optic neuropathy; other ocular, systemic, or neurologic comorbidities that would confound the VF test result were excluded from the investigation.
Patients with were categorized into two groups: (1) normal VF group and (2) abnormal VF group with glaucomatous VF loss.Further, eyes with refractive errors less than -6.0 D were assigned to a highly myopic group.

Red-free fundus photography
All red-free fundus photography had been taken with a simultaneous fundus camera, scanned and saved.One experienced ophthalmologist (HJC) masked to the patients' clinical information reviewed, by double-checking, all of the photos showing sufficient image quality.A single localized RNFL defect (redfree RNFL defect) was defined as a well-outlined, dark, wedge-shaped and not spindle-like defect touching the optic disc border [15].When the photograph was of poor quality to detect the borders, diffuse atrophy or presented multiple RNFL defects in an eye, the case was excluded.The angular width of RNFL defect was measured as previously described (Fig. 1A) [16,17].A circle 3.46 mm in diameter, centred on the optic disc centre, was drawn on red-free photographs.The two borders of each RNFL defect were defined by drawing lines from the centre of the optic disc to each of the points at which the borders of the RNFL defect intersected the circle.The angular width was the angle formed by the two border lines of the RNFL defect.Measurements were obtained with ImageJ software (V.1.48;National Institutes of Health, Bethesda, MD, USA), by two specialists (EB and HJC) each blinded to the patients' clinical history.The average of the two measurements were used as the representative value.

En face imaging of OCTA
The OCTA en face images (6 × 6 mm) centred around the optic disc were scanned using a commercially available OCTA device (AngioPlex).This instrument uses the FastTrac retinal-tracking technology, thereby reducing motion artifacts during the acquisition of OCTA images [18].Poor-quality OCTA images characterized by 1) segmentation failure, 2) motion artifacts leading to irregular vessel pattern or optic disc boundary on enface images, 3) focal weak signal, 4) poor clarity, and/or 5) signal strength < 6 were excluded by a masked reviewer (EB) along with remote review by one other masked reviewer (HJC).En face projections of volumetric scans allowed for visualization of the structural details within the segmented layers of full-thickness retinal scans.The superficial layer of the retina, automatically determined from the boundaries of the internal limiting membrane to the inner plexiform layer, was acquired.
The angular width of the localized RNFL defect determined on the OCT en face image (i.e., en face RNFL defect) was measured as previously described (Fig. 1A) [9,10].En face images were superimposed and aligned to red-free fundus photography according to retinal blood vessel landmarks, using a commercially available software (Photoshop; Adobe, San Jose, CA, USA).The angular width of en face RNFL defects was measured at a distance of 3.46 mm from the centre of the optic disc.The width of the border was measured from the angle made at the centre of the disc by using ImageJ software, as obtained by two specialists (EB and HJC).The average of the two measurements was used for subsequent analyses.When the red-free RNFL defect was not identified on the OCT en face structural image, it was classified as a misidentification with an angular width of zero.

VF measurements
The VF was assessed by the Swedish interactive threshold (SITA) algorithm's standard 24-2 test on the Humphrey Field Analyzer.VF defects were confirmed on two consecutive 6-month reliable tests.Reliable VFs were defined as results with less than 20% fixation loss and less than 25% false-positive and false-negative error rates.The diagnosis was made by at least one examining clinician (HJC) along with remote review by at least one other clinician (EB).Functional outcomes were analysed by mean deviation (MD) and pattern standard deviation (PSD).

Statistical Analysis
The sample size for the study is 256 participants.We calculated that a sample size of 140 patients is required to detect a difference of 0.05 between the two groups, 90% power and using a two sample t-test at the 5% significance level.The subjects' characteristics were compared by independent Student's t-test for normally distributed data and analyzed by chi-squared testing for categorical data.The inter-observer reproducibility of the angular measurements was evaluated by calculating the pairs of intra-class correlation coefficients (ICCs) with their respective confidence intervals (CIs) by two independent examiners (EB and HJC).Pearson correlation analysis and Bland-Altman analysis were used to assess the correlation and agreement, respectively.The correlation between red-free and en face RNFL defects with the functional outcomes was estimated by Pearson correlation analysis using simple linear and second-order polynomial (or quadratic) models.The associations were reported as R 2 (coefficient of determination) with differences between the R 2 calculated using bootstrapping procedures to estimate the 95% CI of the difference in coefficients of determination.For the Bland-Altman plots, the bias with 95% CI was calculated for the angular width of red-free RNFL defect relative to en face RNFL defect.Statistical analyses were performed using SPSS version 23.0 for Windows (SPSS, Inc., Chicago, IL, USA) and R version 3.6.2.(R Project, Vienna, Austria).Pearson correlation coefficients were determined and compared using the Cocor package.A P-value < 0.05 was considered to represent statistical significance.

Demographic and clinical characteristics
In total, 270 glaucomatous eyes of 270 patients meeting the eligibility criteria were enrolled.Among them, 14 were excluded due to poor-quality OCTA images.Eventually, 256 eyes of 256 patients, including 81 highly myopic eyes, met the final entry criteria.In the highly myopic patient subgroup, the SE was −8.02 ± 1.88 D (range, −6.00 to −12.00).The demographic data and a comparison of clinical characteristics between the patients with normal VF and those with abnormal VF are provided in Table 1.
The angular width of en face RNFL defect (range, 0-77.2°) was measured smaller than that of red-free RNFL defect (range, 6.9-86.3°) in 233 eyes (91.0%), and RNFL defect was not identified on OCT en face imaging in 88 eyes (34.4%).Figure 1B provides a scatter plot of the association of angular width of red-free RFNL defect with that of en face RNFL defect.A positive correlation was observed (R 2 = 0.192, P < 0.001).Supplementary Fig. 1A demonstrates the subgroup analysis of highly myopic eyes (R 2 = 0.343, P < 0.001).Figure 1C presents a Bland-Altman plot demonstrating the large discrepancy between the angular width of red-free RNFL defect and that of en face RNFL defect (mean difference: 19.98°, 95% limit of agreement 17.91-22.05°).Supplementary Fig. 1B demonstrates the subgroup analysis of highly myopic eyes (mean difference: 20.73°, 95% limit of agreement 17.34-24.13°).
In Table 2, the angular widths of red-free and en face RNFL defects were significantly larger in abnormal VF patients than in normal VF patients (39.62 ± 16.09 vs. 30.47± 13.49°, P < 0.001 and 24.89 ± 13.75 vs. 1.58 ± 5.12°, P < 0.001, respectively).In highly myopic patients, the angular widths of red-free and en face RNFL defects were also significantly larger in patients with abnormal VF than in those with normal VF (43.64 ± 15.59 vs. 29.38 ± 15.29°, P < 0.001 and 26.48 ± 14.34 vs. 1.62 ± 2.29°, P < 0.001, respectively).En face RNFL defect was observed in only 10 eyes (10.9%) in the normal VF group, whereas 158 eyes (96.3%) in the abnormal VF group (P < 0.001).Among highly myopic eyes, only 2 (8.0%) showed en face RNFL defect in the normal VF group, while all of them showed en face RNFL defect in the abnormal VF group (P < 0.001).

Structure-function association
Figure 2A provides scatter plots of the values of VF parameters against angular width of RNFL defect in the study population.Significant correlations between angular width of red-free RNFL defect and MD (linear R 2 = 0.162, quadratic R 2 = 0.164, all P < 0.001) or PSD (linear R 2 = 0.137, quadratic R 2 = 0.172, all P < 0.001) were identified (Fig. 2A-1, A-2).Comparatively, higher correlations were observed between angular width of en face RNFL defect and MD (linear R 2 = 0.311, quadratic R 2 = 0.320, all Fig. 1 Angular width of retinal nerve fibre layer (RNFL) defect on red-free fundus photography and optical coherence tomography (OCT) en face imaging.A RNFL angle measurement on red-free fundus photography and OCT) en face imaging.A-1 Red-free fundus photography (white box) with a circle centred on the optic disc centre (black line) and borders of the red-free RNFL defect (yellow line).A-2 OCT en face imaging (red box) with borders of the en face RNFL defect (yellow line).B Scatterplot presenting association with width of retinal nerve fibre layer (RNFL) defect angle as detected on red-free fundus photography and optical coherence tomography (OCT) en face imaging.The best-fit linear regression line is shown in blue, and the identity line (where angular width of red-free RNFL defect equals that of en face RNFL defect) is plotted as a red line.A positive correlation was observed (R 2 = 0.192, P < 0.001).C Bland-Altman plots between angular width of RNFL defect as detected on red-free fundus photography and OCT enface imaging.The dashed line in the centre indicates the mean difference between the two images (19.98°).The upper and lower dashed lines are the 95% limits of agreement (17.91 to 22.05°).P < 0.001) or PSD (linear R 2 = 0.372, quadratic R 2 = 0.376, all P < 0.001) (Fig. 2A-3, A-4).The associations between the angular width of en face RNFL defect and VF parameters were significantly stronger than those between the angular width of red-free RNFL defect and VF parameters (P < 0.05 for all pairwise comparisons).

Representative cases
Representative cases of the abnormal VF group with RNFL defect detected on red-free fundus photography versus OCT en face imaging are presented in Fig. 3. Topographically well-matched and different cases of red-free and en face RNFL defects in nonhighly myopic and highly myopic patients are presented.Supplementary Fig. 2 shows two normal VF group cases, both showing visible RNFL defect on red-free fundus photography but no RNFL defect on OCT en face imaging, in non-and highly myopic patients.

DISCUSSION
In the present study, we analysed the angular width of RNFL defect of glaucomatous eyes on red-free fundus photography and OCT en face imaging and compared the strengths of structure-function association of each method.The angular widths between the red-free and en face RNFL defects showed a large discrepancy.En face RNFL defects reflected visual function better than did red-free RNFL defects.Furthermore, this structure-function association of en face RNFL defect was much stronger in highly myopic eyes.
Understanding the structure-function association of glaucoma is important for grading the severity of disease and for understanding the natural history of glaucoma [20].Clinical measurements of structure and function presented a wide variability between individuals at all stages of glaucoma [20].Combining the structure and function tests have improved the diagnostic ability to detect glaucoma [21], but data from large clinical trials have shown structure-function dissociation [22,23].The true extent of damage has often been difficult to ascertain and for the clinician this creates a diagnostic dilemma.Therefore, various investigations have exhibited different methodologies to evaluate the structure-function relationship in glaucoma, such as using the RNFL, inner retinal, GCIPL thickness [24][25][26][27][28][29][30].It has been expected that the integration of structural and functional test results to monitor glaucoma for clinical use [20].To our knowledge, this is the first study in which structure-function association has been evaluated with OCT en face imaging.
Our study showed a relatively weak correlation and proportional biases between the angular widths of red-free and en face RNFL defects, contrary to previous studies [9,10].The discrepancy basically may arise from the two instruments' inherent characteristics in acquiring RNFL images.Red-free fundus photography obtains light reflected from the fundus, blocking the red wavelength by using a green filter.It can maximize the visibility of the RNFL by selectively illuminating the whitish nerve fibre layer against the red background of the retina and choroid [31].By contrast, with OCT en face imaging, RNFL reflectance intensity and thickness is used, and the values of which method vary according to integrity and density [32].Glaucomatous damage not only decreases RNFL thickness but also causes loss of RNFL reflectance.Thus, OCT en face imaging has been reported to offer advantages to detect local glaucomatous damage in great detail [8].
OCT en face imaging presents good agreement with spatial patterns of perimetric defects [7,33,34].Miura et al. studied RNFL defects on widefield OCT en face images, finding that the angular width of RNFL defects correlated with the severity of glaucoma [33].Sakamoto et al. reported that OCT en face images were useful in revealing residual nerve fibre bundles in advancedglaucomatous eyes [34] that deep learning of OCT en face images presented highdiagnostic performance in predicting VF damage severity [35].
Interestingly, in the present study, the structure-function relationship of RNFL defect and VF parameters was stronger in OCT en face imaging than in red-free photography.Our findings suggest that careful observation of OCT en face images may yield better understanding of consequent glaucomatous VF damage.
Evaluating the structural changes associated with glaucoma is challenging in highly myopic eyes.There have been many attempts to overcome these difficulties.First, structural imaging using OCT has improved the diagnostic accuracy of glaucoma [36,37].However, using the thickness of an anatomical layer as a biomarker requires a comparison against a normative database, which fact is a fundamental disadvantage in myopic eyes [38].
Fortunately, OCT en face imaging does not require normative data, and as such, is suitable for highly myopic eyes.Second, structural parameters combined with perimetry has been used.However, highly myopic eyes may be falsely over-diagnosed as glaucoma [39,40].Likely, in the present study, among the 81 highly myopic eyes with red-free RNFL defect, 23 had normal VF without en face RFNL defect.Also, the structure-function relationship correlated more strongly in the OCT en face images than in the red-free fundus photography.Therefore, OCT en face image applicability for glaucoma in cases of high myopia is promising.
Because the nature of the association between the angular parameter of structural and functional measures is not well established, we investigated the relationship between angular Fig. 2 Scatter plots of the values of visual field (VF) parameters against angular width of retinal nerve fibre layer (RNFL) defect.A Scatter plots of the values of VF parameters against angular width of RNFL defect in the total study patients.A-1, A-2 Scatter plots illustrating the linear (bold line) and curvilinear (quadratic fit: dash line) association between angular width of red-free retinal nerve fibre layer (RNFL) defect and mean deviation (MD), pattern standard deviation (PSD).A-3, A-4 Scatter plots showing the linear (bold line) and curvilinear (quadratic fit: dash line) association between angle width of en face RNFL defect and MD, PSD.The associations between the angular width of en face RNFL defect and VF parameters were significantly stronger than those between the angular width of red-free RNFL defect and VF parameters (P < 0.05 for all pairwise comparisons).dB = decibels.*R 2 : Adjusted R 2 from the linear regression model.§ R 2 : Adjusted R 2 from the quadratic regression model.B Scatter plots of the values of VF parameters against angular width of RNFL defect in highly myopic patients.B-1, B-2 Scatter plots illustrating the linear (bold line) and curvilinear (quadratic fit: dash line) association between angular width of red-free retinal nerve fibre layer (RNFL) defect and mean deviation (MD), pattern standard deviation (PSD) in highly myopic eyes.B-3, B-4 Scatter plots showing the linear (bold line) and curvilinear (quadratic fit: dash line) association between angular width of en face RNFL defect and MD, PSD in highly myopic eyes.The associations between the angular width of en face RNFL defect and VF parameters were significantly stronger than those between the angular width of red-free RNFL defect and VF parameters (P < 0.05 for all pairwise comparisons).dB = decibels.*R 2 : Adjusted R 2 from the linear regression model.§ R 2 : Adjusted R 2 from the quadratic regression model.width of RNFL defect and VF damage using a linear and curvilinear model.Regression is the most common and effective method of evaluating the correlation between structural parameters and VF sensitivity [41].It is reported that some sections of SAP are related with RNFL measurements linearly, while other sections show nonlinear association [42,43].On the other hand, other studies support the linear models for relationship between SAP and RNFL measurements [44,45].Therefore, an exact scale is not certain and both linear and logarithmic transformations have been advocated [44,45].Our findings suggest that a quadratic model provides a somewhat better fit to the relationship between angular width of RNFL defect and VF damage than a linear model.From the data set in our study, the low to moderate R 2 values for the structurefunction relationship may seem relatively low, which indicates a weak explanatory power.Glaucoma is a multifactorial disease, and confounding factors such as age, disc morphology, and vascular changes may have influenced the main variable being investigated [46].Also, although a R 2 close to 1 implies an almost perfect relationship between the model and the data, unfortunately, there are no set criteria as to what universally represents a good R 2 value [47].And, as performed in this study, assessing such a statistic is via comparison with another predictive model is preferred [47].This study has possible limitations.First, only eyes with single localized RNFL defects by red-free fundus photography were included.Cases with obscured images, diffuse atrophy or presented multiple RNFL defects were excluded.Hence, in association, we excluded eyes with advanced glaucoma (mean deviation < −12 dB) from the present study.Resultantly, the patient cohort might have been skewed toward mild to moderate glaucoma, which may not represent general glaucomatous eyes.However, by including only eyes with single localized RNFL defects, in which the border of the defect could be clearly defined,

Fig. 3
Fig. 3 Representative cases of retinal nerve fibre layer (RNFL) angle on red-free fundus photography and OCT en face imaging in glaucoma patients.On the left is the localized RNFL defect; in the middle is the en face RNFL defect in the same patient; on the right is the corresponding visual field (VF) loss.A Right eye of non-highly myopic glaucoma patient (spherical equivalent, -1.25 dioptres [D]).The results of the red-free photography and OCT are topographically well-matched.B Left eye of highly myopic glaucoma patient (spherical equivalent, -7.75 D).The results of the red-free photography and OCT are topographically well-matched.C Left eye of non-highly myopic glaucoma patient (spherical equivalent, -1.00 D).The results of the red-free photography and OCT show a topographical difference of angular width.D Right eye of highly myopic glaucoma patient (spherical equivalent, -6.75 D).The results of the red-free photography and OCT show a topographical difference of angular width.

Table 1 .
. And recently, Christopher et al. reported Demographics and clinical characteristics.Mean ± standard deviation.Values with statistical significance are shown in bold.VF Visual field, D Dioptres, IOP Intraocular pressure, C/D Cup-to-disc ratio, dB Decibels.

Table 2 .
Comparison of angular width of retinal nerve fibre layer (RNFL) defect between red-free fundus photography and optical coherence tomography (OCT) en face imaging.Mean ± standard deviation.Values with statistical significance are shown in bold.RNFL Retinal nerve fibre layer, OCT Optical coherence tomography, VF Visual field.