Doppler OCT verifies pulsation-induced anisotropic vessel lumen dynamics in the human retina

Abstract

This study verifies anisotropic vessel pulsatility in human retinas. Doppler optical coherence tomography (OCT) was utilized for dynamic monitoring of peripapillary vessel lumen dynamics. Quantitative analysis consistently revealed larger axial compared to lateral lumen changes, with an axial-to-lateral pulsation change ratio of 1.26 ± 0.08. These findings highlight the anisotropic biomechanical properties of retinal tissues, which contribute to direction-dependent vascular dynamics. Quantitative assessment of anisotropic vascular biomechanics offers promising insights for improving the detection of vascular alterations associated with ocular conditions.

As an extension of the central nervous system, the retina relies on a highly regulated vascular network to meet its high metabolic demands and support visual function. This vascular network dynamically adapts to fluctuating activity levels, ensuring adequate oxygen and nutrient delivery through continuous adjustments in blood flow. The retina’s vasculature is among the most tightly regulated in the body, and unlike other tissues, it lacks significant redundancy, meaning that even minor disruptions in blood flow can lead to dysfunctions and diseases [1]. These vascular responses are governed by neurovascular coupling, the coordinated interaction between neural activities and vascular responses, which is critical for maintaining retinal homeostasis. Disruptions in this balance have been linked to ocular and systemic diseases such as diabetic retinopathy [2], glaucoma [3], and retinal vein occlusion [4], thereby highlighting the clinical significance of retinal vascular dynamics.

Among these vascular dynamics, retinal vessel pulsation, most evident at the optic nerve head (ONH), represents an intrinsic biomechanical response influenced by systemic blood pressure, intraocular pressure, and the mechanical properties of the vessel wall [5–7]. Clinically, these pulsations provide insights into vascular and neurological health, with aberrant pulsation patterns associated with elevated intracranial pressure [8–10] and retinal vein occlusion [11]. Early assessments of pulsation [10,12,13] lacked reproducibility and quantitative rigor. Although subsequent methodologies improved objectivity, they remained constrained to one-dimensional measurements. These early limitations were often driven by technological constraints, such as inadequate imaging resolution and difficulty isolating retinal vessels from surrounding structures. As a result, many early studies relied on indirect measurements or qualitative assessments, which could not capture the full scope of vascular dynamics. Critically, this methodological limitation obscured a fundamental biomechanical property: the anisotropic changes of retinal vessels. These changes arise from the unique mechanical environment of the retina, where the vitreous humor, with its viscoelastic properties, facilitates more pronounced axial expansion. By contrast, lateral expansion is restricted by the stiffer, less compliant retinal tissue, which includes organized collagen structures and cellular layers that resist deformation under pressure [14–17]. This anisotropy reflects the complex interplay between the vascular wall and surrounding tissue, highlighting the directional dependence of biomechanical responses during dynamic physiological conditions. Understanding anisotropic pulsations may provide valuable insights for developing monitoring strategies for diseases where vascular compliance or tissue stiffness plays a role, such as glaucoma or diabetic retinopathy.

Despite established evidence that retinal pulsations align with the cardiac cycle [5], the focus on single-dimensional analysis has offered limited insight into the anisotropic nature of these changes [18–20]. A recent study in a mouse model highlights anisotropic vasodilation in the retina [21], suggesting that intrinsic biomechanical factors underlie direction-dependent expansion differences. However, human studies examining the bi-directional nature of pulsation, an inherently physiological process, remain limited. Addressing this knowledge gap requires imaging capabilities that can capture multi-dimensional vessel dynamics. Although adaptive optics scanning laser ophthalmoscopy (AO-SLO) can measure parabolic flow profiles in mice and humans [22,23], it does not provide the depth-resolved information necessary for fully assessing anisotropic pulsation.

Optical coherence tomography (OCT) provides three-dimensional structural information that extends beyond the capabilities of traditional imaging modalities, such as fundus photography, but does not provide depth-resolved information [24,25]. Although OCT and its functional extension, OCT angiography (OCTA), have improved the visualization of retinal vessels, these modalities remain technically constrained by difficulties in clearly delineating vessels from surrounding tissues for precise vessel lumen quantification [26]. To overcome these limitations while preserving the depth-resolved capability of OCT imaging, Doppler OCT builds on standard OCT imaging by incorporating velocity information and detecting frequency shifts in the back-scattered light, enabling noninvasive, high-resolution assessment of blood flow within the retina [27–30]. By differentiating arteries from veins and measuring pulsation changes along both axial and lateral dimensions, Doppler OCT enables a method for robust quantification of anisotropic pulsation.

This study aims to verify and quantify anisotropic vessel pulsation in healthy human retinas by establishing the axial-to-lateral change ratio as a quantitative measure. Employing Doppler OCT, this study seeks to refine the current understanding of retinal vascular biomechanics and provide evidence of direction-dependent pulsatile behavior. This study was approved by the Institutional Review Board of the University of Illinois Chicago and adhered to the Declaration of Helsinki. Six healthy subjects without a history of ocular and systemic disease ranging in age from 25 to 40 years were recruited from the Lions of Illinois Eye Research Institute of the University of Illinois. Informed consent was provided by each subject before participation in the study.

A custom-designed spectral-domain optical coherence tomography (SD-OCT) system was developed to investigate retinal vessel dynamics (Fig. 1). The system employed a super-luminescent diode (D-840-HP-I, Superlum, Cork, Ireland) with a center wavelength of 840 nm and a bandwidth of 100 nm. Light distribution between the sample and reference arms was managed using a 90:10 fiber coupler (TW850R2A1, Thorlabs, Newton, NJ, USA), with 10% directed to the sample arm. The detection system integrated a spectrometer with a line CMOS camera (Octoplus, Teledyne e2 v, 2048 pixels, 10 × 200 μm) and a transmission grating (1200 lines/mm, Wasatch Photonics, West Logan, UT, USA). The system resolution in the axial and lateral dimensions was theoretically estimated to be 3 μm and 10 μm, respectively. For OCT recording, the illumination power on the cornea was maintained at ~600 μW. All acquired data were saved to a computer hard drive for subsequent post-processing and analysis. The custom-designed OCT system was controlled using lab-built software developed in LabVIEW, providing precise control over imaging parameters and data acquisition.

Fig. 1.

Schematic of the custom-designed optical coherence tomography (OCT) system. SLD, superluminescent diode; PC, polarization controller; CL, collimation lens; lenses, L1, L2, L3, L4, L5, and L6; BS, beam splitter.

Initial volumetric scans (3.5 mm × 3.5 mm) comprising 350 B-scans with 350 A-lines were acquired around the ONH region. Interferograms were processed via fast Fourier transformation for volumetric OCT reconstruction, generating en face OCT images (Fig. 2(A)) with a larger field of view to guide the smaller circular scan at the center. A red-light fixation target minimized involuntary eye movements during imaging. Circular scans (3 mm diameter) were then acquired, with each scan consisting of 2000 A-lines per frame at 10 fps over a 5 s interval (Fig. 2(B)), providing temporal resolution sufficient for pulsatile vessel analysis.

Fig. 2.

Intensity OCT and Doppler OCT analysis of retinal vessels. (A) Representative en face OCT image with a 3.5 mm × 3.5 mm field of view, showing arteries (red: 2, 4, 6, 8, 9, 10) and veins (blue: 1, 3, 5, 7). (B) Circular OCT B-scan image corresponding to the orange circle in (A). (C) Magnified OCT B-scan image corresponding to the blue box in (B) showing the vessel’s circular profile (indicated by white arrows). (D) Corresponding Doppler OCT image. Scale bars: 25 μm.

Before generating both the OCT B-scan and Doppler OCT image from the circular scan, resizing-assisted registration (RAR) [31] was applied using MATLAB R2021a (MathWorks, Natick, MA, USA) to normalize the axial-to-lateral pixel resolution ratio to unity. A magnified OCT B-scan (Fig. 2(C)) from the RAR-normalized circular scan exemplifies this isotropy. The identical 2.5 μm pixel sampling resolution in both lateral and axial dimensions was employed for the following analysis. Doppler OCT processing was then performed on the normalized circular scan data, enabling vessel flow direction mapping within the −π to π phase range, facilitating artery–vein differentiation based on opposite flow directions (Fig. 2(D)). Note that Doppler OCT vessel boundaries (Fig. 2(D)), influenced by processing (e.g., averaging, thresholding, and noise), may differ slightly from anatomical edges (Fig. 2(B)).

From the acquired datasets, 25 vessels demonstrating reliable Doppler signals were selected for quantitative analysis. Vessels included in the analysis were selected based on rigorous criteria to ensure the reliability of Doppler signals. Vessels were included if they exhibited stable, periodic Doppler waveforms, spatial isolation from adjacent vessels, and a signal-to-noise ratio sufficient for consistent detection across the cardiac cycle (typically SNR > 3 dB). Vessels affected by motion artifacts, overlapping signals, or unstable flow patterns were excluded. Furthermore, co-registration between Doppler OCT images and the corresponding structural OCT images was required confirming that the Doppler signal location matched the vessel location in the structural image. This co-registration procedure was essential for reliable vessel differentiation and ensured that only valid signals were analyzed. This process prepared the datasets and selected vessels for analyzing the dynamic pulsatile behavior of retinal vessels. To account for the influence of the angle of incidence light, an angle factor, derived from the vessel’s estimated orientation in 3D OCT data, was used to normalize the axial and lateral diameter change.

Following this preparation, Doppler OCT imaging was utilized to investigate the pulsatile behavior of retinal vessels by analyzing dynamic changes across the cardiac cycle. Figure 3 presents time-lapse Doppler OCT images sampled from a continuous 5 s acquisition period. For optimal visualization of vessel dynamics, Figs. 3(A1) and 3(A2) display 10 sequential time points at 0.1 s intervals, capturing a representative single-second cardiac cycle.

Fig. 3.

Analysis of retinal vessel pulsation using Doppler OCT. (A) Time-lapse Doppler OCT images at 0.1 s intervals over 1 s duration showing pulsatile changes in (A1) artery and (A2) vein selected from the red and blue boundaries in Fig. 2(B). (B) Quantitative analysis showing (B1) center intensity variations of artery and vein, (B2) arterial diameter changes in axial and lateral directions, and (B3) venous diameter changes in axial and lateral directions over three cardiac cycles.

The time-resolved Doppler images (Figs. 3(A1) and 3(A2)) recorded pulsatile changes in arteries and veins, with signal patterns most prominent in the central vessel regions corresponding to peak flow velocity. Quantitative analysis of center intensity variations (Fig. 3(B1)) revealed arterial peak signal intensity levels were approximately double the peak levels found in veins. The temporal analysis identified a measurable millisecond delay between arterial and venous pulsation peaks. This center-based analytical approach was implemented to maximize signal reliability and minimize edge-effect artifacts. Vessel diameter measurements in axial and lateral dimensions (Figs. 3(B2) and 3(B3)) demonstrated distinct temporal patterns across the cardiac cycle, revealing consistent expansion, with axial diameter changes exhibiting greater magnitude compared to lateral dimension for both arteries and veins. The temporal diameter variations showed clear periodicity aligned with cardiac phases, where arterial vessels demonstrated peak expansion during systole followed by diastolic reduction. Quantitative comparison between vessel types revealed that arterial vessels consistently exhibited larger absolute diameter changes relative to venous vessels in both axial and lateral dimensions, with maximum variations coinciding with peak systolic phase. The temporal profiles of these diameter changes maintained reproducibility across multiple cardiac cycles within the 3 s analysis window, confirming the systematic nature of the observed anisotropic vessel behavior. Inter-vessel analysis demonstrated that while the absolute magnitude of diameter changes varied between individual vessels, the fundamental pattern of greater axial compared to lateral expansion remained consistent across the analyzed vessels.

To extend this analysis, Fig. 4 quantifies the axial-to-lateral change ratios for all vessels, providing a quantitative metric for vessel pulsation anisotropy. Figure 4(A) presents a statistical distribution of the axial-to-lateral change ratios for all analyzed vessels through box plot visualization, demonstrating a mean ratio of 1.26 ± 0.08. When analyzed separately by vessel type, the mean ± SD ratio was 1.27 ± 0.08 for arteries and 1.25 ± 0.08 for veins. A Mann–Whitney test indicated no significant difference between vessel types, supporting combined vessel analysis. Statistical analysis of vessel-type dependence incorporated non-parametric testing after confirmation of non-normal distribution through the Shapiro–Wilk analysis (p < 0.05).

Fig. 4.

Axial-to-lateral change ratio analysis for all vessels. (A) Box plot of axial-to-lateral change ratios for all vessels. (B) Scatterplot of vessel diameter versus axial-to-lateral change ratio. Arteries are shown in red and veins in blue.

Correlation-based analysis examining the relationship between vessel diameter and pulsation anisotropy was performed through scatterplot visualization (Fig. 4(B)). The axial-to-lateral change ratio demonstrated consistent values across the observed range of vessel diameters (99.70 ± 12.45 μm), with correlation analysis revealing no significant diameter dependence. The measured ratios remained consistent across all analyzed vessels.

In summary, this study demonstrates anisotropic vessel pulsatility in human retinas through Doppler OCT imaging, revealing a fundamental property of retinal vascular dynamics. The quantitative analysis established an axial-to-lateral change ratio of 1.26 ± 0.08 (Fig. 4(A)), indicating significantly greater vessel lumen changes in the axial direction during cardiac cycles. The observed pulsatile behavior reflects the complex interplay between vessel wall properties and surrounding tissue mechanics. Doppler OCT imaging (Figs. 3(A1) and 3(A2)) clearly visualized this greater axial movement, which can be attributed to the biomechanical environment of retinal vessels.

The greater axial expansion of vessel lumens compared to lateral expansion during pulsation can be attributed to several biomechanical factors. First, the vitreous humor, a gel-like substance, plays a significant role due to its compliance and flexibility, which allows it to accommodate axial vessel expansion with minimal resistance [32]. Its gel phase, primarily composed of collagen and hyaluronan, exhibits viscoelastic properties, including a higher storage modulus than loss modulus, supporting elastic energy retention [33]. Conversely, lateral expansion faces more resistance from the stiffer and rigid retinal layers composed of structured collagen matrices and organized cellular architecture [34]. Additionally, the anisotropic mechanical characteristics of the vessel wall itself may influence this behavior. The axial direction may inherently exhibit greater compliance than the lateral direction, likely due to differences in the organization of smooth muscle cells and connective tissues within the vessel walls [17].

The distinct temporal characteristics observed between arteries and veins align with their known structural differences. Arterial vessels demonstrated larger pulsation amplitudes, particularly during systole (Fig. 3(B1)), reflecting their specialized role in pressure wave propagation. This enhanced arterial response stems from their unique wall composition, characterized by multiple layers of smooth muscle cells and elastic fibers that enable active diameter regulation [34]. Veins, with their thinner walls and reduced elastic fiber content, exhibited smaller pulsation amplitudes, consistent with their primary function as capacitance vessels. As a result, arteries may momentarily exceed the diameter of veins at peak systole due to this elastic expansion, before recoiling and returning to a smaller diameter than veins during diastole. The distinct temporal profiles observed, particularly the difference in diameter curve shapes (Figs. 3(B2) vs 3(B3)), likely reflect these physiological differences; the arterial curve follows the elastic response to the pressure wave, while the venous curve reflects the passive, pressure-sensitive nature of venous walls, influenced by downstream resistance and the dynamics of venous outflow from the eye [18,34].

A particularly significant finding is the stability of the axial-to-lateral change ratio across vessel diameters (99.70 ± 12.45 μm), as demonstrated in Fig. 4(B). This consistency suggests that the anisotropic response represents an intrinsic property of retinal vessel mechanics, likely regulated by the fundamental architecture of the retinal tissue rather than vessel-specific factors. Additionally, the millisecond-scale delay observed between arterial and venous pulsations (Fig. 3(B1)) reflects the progressive transmission of cardiac pressure waves through the retinal vasculature, consistent with established hemodynamic principles and previous findings on arterial and venous pulsation delays [35].

In conclusion, Doppler OCT reveals consistent anisotropic pulsatility in retinal vessels, with greater axial than lateral lumen changes during cardiac cycles (Figs. 3 and 4). This finding emphasizes the intrinsic biomechanical properties of the retinal environment and highlights the need to consider bi-directional vessel dynamics for more accurate vascular assessments.

Data availability.

Data is available upon reasonable request.