Macular Oxygen Saturation in Glaucoma Using Retinal Oximetry of Visible Light Optical Coherence Tomography: A Pilot Study

Abstract

Purpose: A cross-sectional pilot study to compare macular oxygen saturation (sO₂) and associated clinical measurements between normal and glaucoma subjects and to evaluate whether macular sO₂ can be a diagnostic metric for early-stage glaucoma.

Methods: Forty-eight eyes of 35 subjects from three groups were included: normal subjects (16 eyes, 10 subjects), suspect/pre-perimetric glaucoma (GS/PPG) subjects (17 eyes, 12 subjects), and perimetric glaucoma (PG) subjects (15 eyes, 13 subjects). We performed retinal oximetry of visible light optical coherence tomography (VIS-OCT) in macular vessels, with 512 × 256 sampling points over a 5 × 5 mm² area. Zeiss Cirrus OCT scans and a 24-2 visual field test (VFT) were conducted. Statistical analysis was conducted.

Results: Significant differences were observed among the three groups for all VIS-OCT, Zeiss OCT, and VFT variables. As glaucoma severity increased, macular AsO₂ (arterial sO₂) and A-V sO₂ (arteriovenous sO₂ difference) decreased, whereas macular VsO2 (venous sO₂) increased. Macular AsO₂ and A-V sO₂ were found to be statistically correlated with ganglion cell layer + inner plexiform layer (GCL+IPL) and circumpapillary retinal nerve fiber layer in all eyes, as well as in PG eyes. Within the PG group, a dominant correlation between AsO₂ and ganglion cell layer + inner plexiform layer was observed in the more damaged lower hemifield.

Conclusions: Glaucoma subjects showed altered macular sO₂, indicating reduced oxygen consumption. The sO₂ measured by VIS-OCT could be a potential metric for early glaucoma diagnosis.

Translational Relevance: This study shows macular sO₂ measurements via VIS-OCT could bridge advanced imaging technology and clinical glaucoma detection.

Introduction

Glaucoma is an optical neuropathy, causing irreversible blindness and impacting millions worldwide.^1,2 The clinical hallmark of glaucoma is the loss of retinal ganglion cells (RGCs) and nerve fiber layer (RNFL), leading to chronic vision deterioration. Vision loss in glaucoma is permanent and irreversible, emphasizing the importance of early detection for prevention and clinical management.^3–6 

Although the initial insult occurs at optic nerve head (ONH), it retrogradely damages the axon and soma of RGCs.⁷ Because the macula contains >30% of total RGCs in retina,^8–10 changes associated with RGC damage may occur and be detected in macular region. Indeed, evidence supported that early stages of glaucoma may involve macular RGC loss that warrants close attention.^8,11 Thinning of the macular ganglion cell complex is associated with glaucoma.¹² Optical coherence tomography angiography found reduced macular superficial capillary density in glaucomatous eyes, and the rate of capillary density can indicate glaucoma worsening in some cases.^13–15 Using a 10-2 visual field test (VFT) with a denser 2° sampling density, a recent study showed central visual damage in more than 30% of ocular hypertensive and glaucomatous eyes, whereas the conventional 24-2 VFT appeared normal.¹⁶ 

The retina is a tissue with one of the highest levels of oxygen consumption, and RGCs are the major user in the inner retina.^17,18 The macular region is perfused by the parafoveal arterioles and venules extended from the superior and inferior arcades, in an alternating pattern.¹⁹ Similar to the major vessels near ONH, the parafoveal arterioles and venules maintain a balanced blood inflow and outflow, and a localized oxygen supply-demand equilibrium. Because of the abundance of RGCs at the macula, we hypothesize that macular inner retina oxygen extraction can assess local RGC health, and a reduced oxygen extraction indicates the damage or loss of RGCs, by measuring blood oxygen saturation (sO₂) from parafoveal arterioles and venules. 

The measurement of macular sO₂ from small arterioles and venules from human eyes remains a challenging task. Existing retinal oximetry uses dual- or multi-wavelength fundus imaging or hyperspectral scanning laser ophthalmoscopy lack depth-resolving capability,^20–23 which is confounded by pigmentation and central line reflection. They also lack resolution to provide detailed assessment of small macular vessels. To overcome these challenges, visible light optical coherence tomography (VIS-OCT) was used in this study. VIS-OCT is an emerging retinal imaging method that uses shorter wavelengths of visible light instead of conventional near-infrared (NIR) light, resulting in much higher axial resolution and a significant absorption contrast between oxygenated and deoxygenated hemoglobin.^24–26 The differential absorption contrast with VIS-OCT can then be leveraged to quantify sO₂ in human eyes.^19,27,28 

Here, the goal of this pilot study is to use VIS-OCT retinal oximetry to measure macular sO₂ in normal, suspect/pre-perimetric glaucoma (GS/PPG) and perimetric glaucoma (PG) groups; to evaluate the feasibility of sO₂ as an early biomarker to differentiate among these three groups; and to explore the association between sO₂ and glaucoma-represented thickness parameters of ganglion cell layer + inner plexiform layer (GCL + IPL) and circumpapillary retinal nerve fiber layer (cpRNFL). 

Method

Human Subjects

The Institutional Review Board of Boston Medical Center reviewed and approved this study, ensuring compliance with the Health Insurance Portability and Accountability Act. The study took place from March 2019 to January 2020. All the subjects were provided with the tenets of Declaration of Helsinki, and written informed consent was obtained from each participant. We recruited the subjects in the control group through the Boston Medical Center Optometry clinics and the clinical subjects from the Boston Medical Center Ophthalmology clinic during their standard of care visit. 

Inclusion and Exclusion Criterion

We included subjects over 40 years old and with visual acuity better than 20/40 after optical correction. For the experimental group, we only included the patients diagnosed with primary open-angle glaucoma or potential primary open-angle glaucoma, excluding other ocular conditions such as primary angle closure glaucoma, intraocular surgeries (except for uncomplicated cataract surgery), history of diabetic retinopathy, vascular occlusion, macular degeneration, macular edema, hereditary retinal degeneration, uveitis, traumatic glaucoma, and other retinal conditions. We excluded subjects from both groups as a result of fixation failure and low image quality, which rendered them unsuitable for segmentation. Subjects with severe cataracts graded higher than 2+ were also excluded, considering the sensitivity of visible light to cataract and lens changes. 

Clinical Examination

Subjects underwent tonometry, stereoscopic optic disc assessment and clinical OCT imaging (Cirrus, Zeiss, Jena, Germany). We recorded the quantitative results including cpRNFL and GCL + IPL, cup-to-disc ratio (CDR) and vertical cup-to-disc ratio (VCDR) from Cirrus OCT scans. For all the clinical subjects, central 24-2 threshold VFTs were conducted with mean deviation (MD) and pattern standard deviation (PSD). We evaluated the cataracts using the Lens Opacification System II based on color and opalescence using a four-point grading system with an increasing number consistent with increasing maturity. After clinical and ophthalmic examinations, a trained technician imaged the subjects using the dual-channel VIS-OCT system on both eyes, if eligible. The pupils were dilated by phenylephrine or tropicamide for standard-of-care purposes or to generate the highest quality images. Appropriate medication administration was determined by a review of allergies, ocular structure, intraocular pressure (IOP), baseline pupil size/reactivity, and previous dilation experiences. 

We acquired the fovea-centered images using a custom-built dual-channel VIS-OCT system²⁹ with a raster scan of 512 A-lines by 256 B-scans covering an area of 5 × 5 mm². The illumination wavelengths were 545 to 580 nm (35 nm bandwidth) for the visible channel, where the hemoglobin absorption peaks and the contrast between oxygenated and deoxygenated hemoglobin is strong (i.e., isosbestic point at 545, 571, and 584 nm). In addition, a narrower bandwidth improves the spectral power density with the same averaged power, and therefore the spectral signal-to-noise ratio for sO₂ calculation with the tradeoff of a lower axial resolution. NIR channel uses 800 to 880 nm bandwidth. The illumination power on pupil was less than 0.25 mW (VIS) and 0.9 mW (NIR), respectively, meeting the safety standards of the ANSI for ophthalmic instruments. Details of the calculation can be seen in the Supplemental Material from our previous publication.³⁰ We used a tunable lens to correct the spherical refractive error. We used the fellow eye for fixation with an external target. The start of the imaging used the NIR channel for alignment and focus tuning. After confirming centered location at the macula with the best preview image quality, we initiated the dual-channel acquisition. The A-line rate of camera was 50 kHz with an exposure time of 19.1 microseconds. The total acquisition time for one raster scan was 2.62 seconds. 

Image Processing and sO₂ Calculation

After acquisition, volumetric three-dimensional data from both channels were produced by DC spectrum removal, k-space resampling, dispersion compensation, and fast Fourier transform. Detailed processing steps are described in the past work.¹⁹ Because of more power density in the NIR channel, we first used NIR data to segment the retinal pigment epithelium (RPE) layer by detecting the location of maximum intensity in each A-line and then detected the location of the maximum gradient above the RPE layer for internal limiting membrane (ILM) boundary. We used outlier detection and third-order polynomial curve fitting to smooth the segmentation of the RPE and ILM and then registered them in VIS data. We generated the en face vessel map by averaging the space between the RPE and 20 µm above RPE for maximum contrast and then manually chose the vessels for vessel mask of region of interest (ROI). The ROI for vessel selection were primarily the branch arterioles and venules within the macular area, ensuring consistency in locating the vessel bottom for subsequent sO₂ calculations. Vessels were classified into hierarchical levels based on their branching structure from ONH: grandparent vessels (large arteries and veins emerging directly from the ONH), parent vessels (first-level branches), and children and grandchildren vessels (progressively smaller branches). For sO₂ analysis, only vessels within the macular region (parafoveal and macular circles in Fig. 1) were included. Within the macula, larger vessels—mainly grandparent vessels—were excluded because they also perfuse inner retina tissue outside macula. Additionally, only vessel segments containing more than 150 A-lines (empirically determined) were included to ensure measurement reliability. We assigned the vessels into arterioles and venules based on their characteristic alternating pattern in the macular region, and cross-reference to the anatomical branching if available. 

Figure 1.

 

The sO₂ maps and sO₂ calculation. (a–c) Representative sO₂ maps of the macula from right, left, and right eyes in Normal, GS/PPG, and PG groups, showing spatial oxygen saturation distributions. (d, e) Linear square fitting applied at selected wavelength points to calculate sO₂ values of arterioles and venules.

Using a short-time Fourier transform, 11 Gaussian windows were swept along the interferogram for a four-dimensional dataset I (x,  y,  z,  λ). A-lines within vessel ROIs were averaged for the spectrum I (z,  λ), which was further normalized by an averaged spectrum from the nonvascular RNFL. With the known ILM boundary, we located the vessel bottom¹⁹ and averaged the spectrum in depth from 5 µm above the vessel bottom to 10 µm below to generate a single spectrum for each vessel ROI. 

As shown in Figure 1, a least-square fitting on the extracted spectra calculates sO₂ for each vessel ROI using the algorithm below:  

\begin{eqnarray} && I(s{{O}_2}|\lambda ,z)\nonumber\\ && \quad = {{I}_0}(\lambda )\sqrt {{{R}_0}r\left( \lambda \right)} {{e}^{ - [s{{O}_2} \times {{\mu }_{Hb{{O}_2}}}\left( \lambda \right) + \left( {1 - s{{O}_2}) \times {{\mu }_{Hb}}\left( \lambda \right)} \right]z}}\quad \end{eqnarray}

(1)

where I₀(λ) is the spectrum of light source; R₀ is an assumed constant for the reflectance of reference arm; r(λ) (dimensionless) is the reflectance at the vessel wall, modeled by a power law r(λ) = Aλ^−α, with A being a dimensionless constant and α modeling the decaying scattering spectrum from the vessel wall. The optical attenuation coefficient μ is determined by the attenuation coefficients of absorption μ_a and scattering μ_s, where μ(λ) = μ_a(λ) + Wμ_s(λ). W, was equivalent to 0.2 in this study, is the scaling factor for the scattering coefficient. After obtaining the sO₂ for all vessels, we classified them into the sO₂ of arterioles (AsO₂) and venules (VsO₂) based on the alternation pattern and values. We defined the sO₂ difference between arterioles and venues (A-V sO₂ = AsO₂ − VsO₂) and oxygen extraction (OE = [AsO₂ – VsO₂]/AsO₂ ∙ 100%) metrics for further investigation. 

Key steps involve manual intervention: (1) vessel segmentation from en face projection to produce vessel ROI; (2) inspection of the vessel bottom identification; and (3) arteriole and venule assignment. 

Study Group Definition

We defined a normal eye as one with a normal-appearing ONH, assessed by a stereoscopic optic disc photograph at the point of care (subjective assessment of vertical cup-disc ratio), and IOP ≤ 22 mm Hg. We defined the GS/PPG as an optic disc potentially presenting glaucomatous optical neuropathy after a stereoscopic optic nerve examination. The visual field was normal by the 24-2 VFT threshold test, and the glaucoma hemifield test was within normal range or borderline. The index of pattern standard deviation was less than 5%. We defined PG as an optic disc compatible with glaucoma. The subjects exhibited an abnormal visual field, including a result for the glaucoma hemifield test outside the normal limit or the index of pattern standard deviation was less than 0.5%. 

Statistical Analysis

We used various statistical methods tailored to the data type, sample size, and specific analytical needs of each comparison. 

Group Comparisons

For categorical variables in Tables 1 and 2, we used the Fisher-Freeman-Halton exact test when sample sizes were small or when expected frequencies in any contingency table cell were less than five, because this test provides accurate P values under these conditions. For larger samples meeting χ² test assumptions, specifically an expected frequency of at least five in each cell, the χ² test was applied. For continuous variables, we performed one-way analysis of variance (ANOVA) across the three groups (normal, GS/PPG, and PG) to assess group differences. Recognizing the potential correlation between two eyes from the same subject, we applied linear mixed models (LMM) to estimate adjusted means and standard errors, followed by Tukey's method for multiple comparisons within the three groups, as shown in Figure 2. This approach allowed us to account for within-subject variability and obtain more robust comparisons. We used the Mann-Whitney U test to compare two independent groups when the data distribution was unknown or when normality could not be assumed in Figure 3b and Supplementary Figure S2. 

Table 1.

 

Demographic of Subjects

 

Table 2.

 

Characteristics of Ocular Measurements From Eyes Among Three Groups

 

Figure 2.

 

Statistical comparison among normal, GS/PPG, and PG groups using adjusted mean values and standard errors derived from a linear mixed model (LMM). (a–d) Scatter points represent individual measurements of AsO₂, VsO₂, A-V sO₂, and OE in arterioles and venules across the three groups, showing trends in oxygen saturation parameters with disease progression. (e, f) Scatter points for GCL + IPL and cpRNFL thickness measurements obtained from Zeiss OCT show structural changes with glaucoma severity. Adjusted mean values and significance levels (*P < 0.05, **P < 0.01, ***P < 0.001) are noted, with Tukey's method used for multiple comparisons within groups.

Figure 3.

 

Hemifield analysis in PG subjects. (a) The PG right eye showing the deviation map, visual field map, and regional GCL+IPL thickness. (b) Comparison of MD values between the lower and upper hemifields of PG subjects (**P < 0.01). (c–f) Scatter plots demonstrate the correlations between AsO₂, VsO₂, and GCL+IPL thickness in the lower and upper hemifields. Each scatter point represents an individual measurement, with correlation coefficients and P values noted to indicate the strength and significance of observed relationship.

Correlation Analysis

To evaluate the relationships between continuous variables, we used Spearman's rank correlation in Figures 3c–f, Tables 3 and 4, and Figures S1b and S1c. Spearman's correlation was selected because it is a nonparametric test suitable for data that may not meet normality assumptions, and it measures the strength and direction of monotonic relationships between variables. This analysis provided correlation coefficients and P values, helping us interpret associations between parameters. Additionally, we applied LMM to account for potential correlations between two eyes from the same subject, as presented in Supplemental Tables S1 and S2. 

Diagnostic Accuracy

To assess the diagnostic accuracy of individual parameters, we performed receiver operating characteristic (ROC) curve analysis (Table 5) to evaluate the sensitivity and specificity of diagnostic tests by calculating the area under the curve (AUC). ROC curves were generated from logistic regression models to determine the discriminative power of each parameter, with AUC values interpreted as follows: values between 0.5 and 0.7 indicate low accuracy, 0.7 to 0.9 indicate moderate accuracy, and values above 0.9 indicate high diagnostic accuracy. Parameters with significant diagnostic value (P < 0.05) are highlighted in bold. 

In all analyses, a P value < 0.05 was considered statistically significant. 

Results

Demographics

A total of 48 eyes from 35 subjects were included in the analysis: 16 eyes from 10 normal subjects, 17 eyes from 12 GS/PPG subjects, and 15 eyes from 13 PG subjects. Subjects with severe cataracts or fixation failure, which led to incomplete en face macular images, were excluded due to the inability to segment retinal layers for sO₂ analysis. 

As shown in Table 1, no significant demographic differences were observed among the groups in terms of eye side, age, gender, race, or ethnicity. The χ² tests were used for eye side and gender, whereas the Fisher-Freeman-Halton exact test was applied for race and ethnicity because of smaller sample sizes. One-way ANOVA was used to evaluate age differences, confirming demographic comparability across the groups. 

Key Measurements

Ocular measurements across the three groups are summarized in Table 2, revealing significant differences consistent with disease progression. 

In VFT, both 24-2 MD and 24-2 PSD showed significant deterioration between the GS/PPG and PG groups, indicating worsening visual function with disease severity (P < 0.001 for both, using t-tests). 

For structural parameters measured by Zeiss OCT, significant differences were observed across the three groups using one-way ANOVA. GCL + IPL and cpRNFL thickness decreased as disease severity increased (P < 0.001 for both). Additionally, averaged CDR and VCDR significantly increased from normal to glaucoma groups (P < 0.001 for both), reflecting retinal thinning and optic disc changes associated with disease progression. 

Oxygen saturation parameters measured by VIS-OCT also showed significant differences across groups using one-way ANOVA. AsO₂ decreased with disease severity (P = 0.017), whereas VsO₂ increased from normal to glaucoma groups (P = 0.009). A-V sO₂ and OE, both representing oxygen extraction, declined significantly across the groups (P < 0.001 for both), suggesting reduced macular oxygen consumption in glaucomatous eyes. These observed patterns in oxygen saturation parameters may correlate with functional and structural decline as glaucoma progresses, although formal trend testing was not conducted. For other ophthalmic measurements, a significant difference was observed in cataract presence among the groups (P = 0.013, Fisher-Freeman-Halton exact test), although pseudophakic lens presence and intraocular pressure (IOP) were not significantly different. 

Figure 2 illustrates the distribution of VIS-OCT measurements across the three groups (normal, GS/PPG, and PG), including AsO₂, VsO₂, A-V sO₂, OE, GCL + IPL, and cpRNFL. Adjusted mean values and standard errors were derived using an LMM to account for potential associations between both eyes from the same subject, with Tukey's method applied for multiple comparisons within groups. 

The adjusted mean of AsO₂ shows a progressive decrease from normal to GS/PPG and further to PG, with significant differences observed between normal and GS/PPG (P < 0.05), indicating a reduction in arterial oxygen saturation as disease severity increases. VsO₂ exhibited an increase from normal to the glaucoma groups, with significant differences observed between normal and GS/PPG and between normal and PG (P < 0.05 for both). A-V sO₂ and OE, representing oxygen extraction, demonstrated a decreasing pattern with increasing disease severity. Significant differences were noted between normal and GS/PPG (P < 0.01) and between normal and PG (P < 0.001), suggesting reduced macular oxygen consumption as glaucoma progresses. For Cirrus OCT thickness measurements, both GCL + IPL and cpRNFL decreased with disease severity. Significant differences were observed from normal to PG (P < 0.001 for both) and from GS/PPG to PG (P < 0.05 for GCL + IPL, P < 0.01 for cpRNFL), reflecting structural thinning associated with advanced disease stages. 

Correlations of Key Parameters

We performed Spearman correlation analysis among key parameters across the three groups, focusing on the relationships between sO₂ variables (AsO₂, A-V sO₂, and OE) and structural parameters (GCL + IPL and cpRNFL), as shown in Table 3. Significant correlations were observed between sO₂ metrics (except VsO₂) and structural measurements. GCL + IPL showed significant positive correlations with AsO₂ (r = 0.341, P = 0.019), A-V sO₂ (r = 0.356, P = 0.014), and OE (r = 0.316, P = 0.031). Similarly, cpRNFL was significantly correlated with AsO₂ (r = 0.339, P = 0.019), A-V sO₂ (r = 0.363, P = 0.012), and OE (r = 0.304, P = 0.038). These findings indicate that lower oxygen extraction, as indicated by A-V sO₂ and OE, is associated with structural thinning in GCL + IPL and cpRNFL, potentially reflecting the impact of glaucomatous damage on retinal oxygenation and structure. 

Table 3.

 

Correlation Coefficients With P Values in Parentheses for Key Measurements Among Three Groups

 

We further investigated the correlations within the PG group, as shown in Table 4. Compared to the results across all groups in Table 3, GCL + IPL and cpRNFL continued to show significant correlations with key sO2 parameters, specifically AsO₂ and A-V sO₂. Within the PG group, GCL + IPL was significantly correlated with AsO₂ (r = 0.596, P = 0.024) and A-V sO₂ (r = 0.546, P = 0.044). Similarly, cpRNFL showed significant correlations with AsO₂ (r = 0.581, P = 0.029) and A-V sO₂ (r = 0.568, P = 0.034). These findings indicate that, even within the advanced stage of glaucoma, higher oxygen saturation levels and oxygen extraction remain associated with structural thinning. 

Table 4.

 

Correlation Coefficients With P Values in Parentheses for Key Measurements in PG

 

To further assess the relationship between paired parameters while accounting for the potential correlation between bilateral eyes from the same individual, we employed a linear mixed model (LMM). This analysis was conducted across all subjects (Supplementary Table S1) and specifically within the PG group (Supplementary Table S2), confirming a significant relationship between GCL + IPL thickness and sO₂ parameters derived from A-V sO₂ and OE. These results align with the correlation analysis and support the association between structural thinning and oxygen saturation changes in advanced glaucoma. 

Hemifield and Quadrant Analysis

We examined macular sO₂ by hemifields within PG eyes, as visual damage is known to be more severe in the lower hemifield compared to the upper hemifield (Figs. 3a, 3b), with significantly lower MD values observed in the lower hemifield (P < 0.01). This pattern aligns with the characteristic glaucomatous damage distribution. When comparing macular AsO₂ with macular GCL + IPL thickness, Figure 3c shows a positive correlation trend between AsO₂ and GCL + IPL in the lower hemifield, with a P value = 0.063. Although this did not reach conventional statistical significance (P < 0.05), the near significance suggests a potential association that may support our conclusion regarding AsO₂’s relationship with macular RGC loss in regions with confirmed visual field damage. 

Exploring the quadrant-specific analysis within the PG group (Supplementary Fig. S1), we assessed the correlation between quadrant-specific GCL+IPL thickness and AsO₂ and VsO₂. Notably, a significant correlation was observed between GCL+IPL thickness and AsO₂ in the nasal-inferior and inferior quadrants (P = 0.028 and P = 0.046). This finding suggests that AsO₂ is closely related to structural thinning in specific regions of the macula, particularly in areas more susceptible to glaucomatous damage. No significant correlations were observed between GCL + IPL thickness and VsO₂ in any quadrant. We compared AsO₂ and VsO₂ between the upper and lower hemifields in the PG group, and AsO₂ has larger difference than VsO₂, consistent with the above hemifield analysis where AsO₂ is more correlated with GCC thinning in more damaged upper hemifield in Supplementary Figure S2. 

ROC Analysis for Diagnostic Accuracy

Table 5 illustrates the AUC results of the ROC analysis, identifying the most effective parameters for distinguishing among the three groups. All sO₂ parameters performed better than GCL + IPL and cpRNFL in distinguishing between normal and GS/PPG, as well as between normal and the combined group of GS/PPG + PG, with OE showing the highest AUC (0.849 for normal vs. GS/PPG and 0.855 for normal vs. GS/PPG + PG). In differentiating GS/PPG from PG, cpRNFL achieved the highest AUC (0.775), indicating its effectiveness in distinguishing more advanced stages of glaucoma. 

Table 5.

 

AUC by ROC Analysis

 

Discussion

This is the first study to investigate macular sO₂ of glaucoma subjects using the retinal oximetry of dual-channel VIS-OCT. VIS-OCT uniquely allows scanning of the macular arterioles and venules to calculate macular sO₂ without the confounding signals from other layers that plague traditional fundus-based oximetry. 

RGCs are most abundant in macula, and that they are the major energy consumer in inner retina. Unmyelinated RGC axons up to the lamina cribrosa require significant amounts of energy to generate and propagate action potentials,³¹ and RGC somata have high energy demand to actively synthesize transmitters, transport cargo, and maintain ion balance across cell membranes.³² These tremendous energy needs are subserved by the dense mitochondria in RGC somas in the macular region and within the IPL where synapses are formed.^33,34 Our finding suggests that early RGC/RNFL loss may reduce the energy demand, which is manifested with reduction of oxygen extraction in the macular region, correlated with the thinning of GCL + IPL + RNFL. 

We observe variations in sO₂ among different arterioles and venules in the same eyes. In normal retinas, the variability is more likely determined by the measurement accuracy, with the limited number of A-lines per branch. Both the spectral signal and algorithm can contribute to the measurement variation. Vessel averaging effectively reduces variability. Our previous study achieved a coefficient of variation (CV) < 3% within a single day and <5% across five weeks with ∼4× more A-lines per vessel than the current study.³⁵ For pathological eyes, the difference from individual branches may be also due to pathological changes. Nonetheless, the empirical vessel averaging within an eye or a hemisphere shows correlation between macular sO₂ markers with the severity of glaucoma. In any case, the improvements in imaging devices and algorithms will improve the accuracy of VIS-OCT macular oximetry, ultimately leading to better clinical utility. 

We found macular sO₂ is correlated with severity of glaucoma and performs better at separating GS/PPG from normal eyes than GCL+NFL and cpRNFL thinning. The declination of AsO₂ in more severe stage of glaucoma is interesting, which could indicate pathological role of vascular function or merely a consequence of RGC loss and compromise of neurovascular coupling. Further studies are required to delineate the causal relationship. Within the PG group, AsO₂ is significantly correlated with GCL + NFL and cpRNFL thicknesses, which is attributable by the more severely damaged lower hemifield. This correlation analysis strongly indicates that macular sO₂ is associated with glaucomatous macular tissue loss. Importantly, macular VsO₂ and A-V sO₂ have significant differences when comparing normal with GS/PPG eyes, even before detectable thinning and visual field damage. 

All existing literature regarding sO₂ in glaucoma reported global measurements from the major vessels in the parapapillary region around the ONH, which is difficult to extrapolate to macular sO₂ in this presented study.^36–43 Nonetheless, those reports showed rather consistent AsO₂ in major retinal arterioles, or in rare cases, higher AsO₂ in PG subjects. Although the inclusion criteria and subject population differ, those reports also showed decreased global A-V sO₂ in glaucoma, underlay by reduced metabolic demand with RGCs loss. Remarkably, we observed a similar reduction of A-V sO₂, as well as OE at a more localized macular region. Given that RGCs are most abundant in macula, and that they are the major energy consumer in inner retina, our finding suggests that early RGC/RNFL damage may be manifested with reduction of oxygen metabolism in the macular region, correlated with the thinning of GCL + IPL + RNFL. It is worth noting that, contrary to an unchanged (or slightly increased) global AsO₂ in previous reports,^37–43 we found that macular AsO₂ significantly declined with glaucoma severity among three groups. This difference is primarily due to the distinct measurement locations and methodologies. 

Both AsO₂ and A-V sO₂ are significantly correlated with thicknesses of GCL + NFL and cpRNFL either among three groups or within PG eyes. In addition, in PG eyes, the correlation is primarily driven by the more severely damaged lower hemifield. To our surprise, we didn't find a significant correlation between VsO₂ and GCL+NFL or cpRNFL, whether among three groups or within PG eyes. Upon closer examination of the data, VsO₂ increases significantly from normal to GS/PPG, but remains relatively consistent between two more severer groups. We speculate that VsO₂ is more significantly impacted in early stage of tissue atrophy in glaucoma, while the AsO₂ continues to decline along with the progressing severity. A more comprehensive dataset including macular blood flow would further elucidate the speculation. 

We found no significant correlation between global MD and any macular sO₂ parameters or GCL + IPL/cpRNFL. We note that 24-2 VFT only has four measurement points within a 25° center viewing angle in the macula, thus undersampling the macular region, which may lead to large variations.¹⁶ Also, because the RGCs are highly redundant in macula up to 30% of RGC, loss can proceed to visual field damage.¹⁰ Therefore, it is not totally surprising that correlations between MD and other variables are not statistically significant within this study population. 

VIS-OCT holds significant promise for advancing glaucoma diagnostics and patient care. Beyond its ability to measure macular sO₂, which is key for early detection and monitoring of glaucoma, VIS-OCT can provide several other advantages. Its ultrahigh axial resolution allows for detailed visualization of retinal microstructures, particularly the sub-band changes in the outer retina.⁴⁴ Additionally, VIS-OCT can assess the reflectivity of RNFL, which correlates with glaucoma severity and provides another sensitive marker for early disease detection.³⁰ The technology also facilitates fibergraphy, allowing for the visualization and quantification of RGC axon bundles, which are essential for understanding disease progression.⁴⁵ These capabilities make VIS-OCT a comprehensive tool for not only diagnosing and monitoring glaucoma but also for potentially improving the overall management in clinical practice. 

There were several limitations of this study that we would like to address in future studies. First, the cohort size was limited. A larger subject population would be beneficial for multivariable statistical analysis. Second, the implementation of visible light can cause discomfort to some subjects, and it is more susceptible to cataracts and aging eyes. To address this, we will further optimize the imaging device including eye tracking and automatic focusing, which will significantly reduce exposure to visible light. Third, the current data processing used manual segmentations for selecting vessel ROIs, as well as assignment of arterioles and venules based on experience. Potential bias and subjectivity can be introduced by manual segmentation and vessel assignment. To mitigate this issue, we standardized our approach for all subjects among three groups. This standardized approach helps to minimize bias and improve the reliability of our findings. Future studies will incorporate automated segmentation algorithms and vessel tracing to streamline the data processing, further reducing potential bias and enhance reproducibility. Finally, to balance the image depth range and signal to noise ratio, the device used in this study has a relatively short visible light bandwidth of 35 nm. Our new-generation VIS-OCT device improved the resolution with wider wavelength bands and expanded the working range.⁴⁶ With more precise extracted spectra, it will help increase the calculation accuracy of sO₂ in future studies. 

In conclusion, we report the first study of VIS-OCT to investigate the macular sO₂ in glaucoma subjects. With its high resolution in both axial and lateral directions, VIS-OCT is a powerful tool to characterize macular sO₂ and identify the significant differences between normal, GS/PPG and PG subjects. The measurement of sO₂ can potentially act as a marker to provide early diagnosis and monitor the progress of glaucoma disease. 

Acknowledgments

Supported by NIH funding R01NS108464, R01EY032163, and R01EY034607. 

Disclosure: J. Wang, None; N. Sadlak, None; M.G. Fiorello, None; M. Desai, None; J. Yi, None