Open Access
Add to BibSonomy
Abstract
A new method based on polarization-sensitive optical coherence tomography (PS-OCT) is introduced to determine the polarization properties of human retinal vessel walls, in vivo. Measurements were obtained near the optic nerve head of three healthy human subjects. The double pass phase retardation per unit depth (DPPR/UD), which is proportional to the birefringence, is higher in artery walls, presumably because of the presence of muscle tissue. Measurements in surrounding retinal nerve fiber layer tissue yielded lower DPPR/UD values, suggesting that the retinal vessel wall tissue near the optic nerve is not covered by retinal nerve fiber layer tissue (0.43°/µm vs. 0.77°/µm, respectively). Measurements were obtained from multiple artery-vein pairs, to quantify the different polarization properties. Measurements were taken along a section of the vessel wall, with changes in DPPR/UD up to 15%, while the vessel wall thickness remained relatively constant. A stationary scan pattern was applied to determine the influence of involuntary eye motion on the measurement, which was significant. Measurements were also analyzed by two examiners, with high inter-observer agreement. The measurement repeatability was determined with measurements that were acquired during multiple visits. An improvement in accuracy can be achieved with an ultra-broad-bandwidth PS-OCT system since it will provide more data points in-depth, which reduces the influence of discretization and helps to facilitate better fitting of the birefringence data.
© 2021 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Untreated vascular disorders increase the risk of health diseases that affect the heart, brain, kidneys and arteries [1]. The analysis of blood vessel properties, in vivo, can have a critical role in the diagnosis and monitoring of vascular disorders [2]. Optical imaging methods have a resolution advantage over magnetic resonance imaging (MRI) and do not use ionizing radiation. The eye and its retina can readily be observed with optical imaging techniques, in vivo. Since the blood vessels in the eye and the heart share common vasculature properties, the retinal vasculature may serve as a proxy to non-invasively assess cardiovascular conditions with optical imaging technology [3]. A perhaps often overlooked imaging target that could be useful for the early diagnosis of vascular diseases are the walls of a blood vessel. They consist of organized fibrous connective tissue and since microvasculature disorders disrupt the organization of the fibrous structure, the assessment of organizational loss offers an opportunity to diagnose vascular diseases. The same method can also be used to diagnose multiple retinal diseases [4–6].
OCT is an interferometry-based imaging method that can non-invasively obtain cross sectional tomographic images of the eye and retina [7]. OCT images are generated by the detection of the interference between backscattered light returning from the sample and reference arms of an interferometer. The method uses safe levels of near infrared light as the light penetrates easily through the tissue with sufficient back scattering for images with a high signal-to-noise ratio (SNR).
Polarization-sensitive OCT (PS-OCT) uses polarized light to retrieve the depth-resolved polarization properties of tissue, such as birefringence, the degree of polarization uniformity and (changes in) the fast axis orientation [8,9]. It has been applied in the human retina to retrieve retinal nerve fiber birefringence [10,11], the degree of polarization uniformity of the retinal pigment epithelium (RPE) [12,13], the retardance induced by the Henle fiber layer [5,14] and the retardance induced by the RPE-Bruch’s membrane complex [15].
Aging, macular disorders and hypertension can disrupt the organization of fibrous structures in the vessel wall, and we hypothesize that this should lead to a reduction in tissue birefringence [16,17]. The retardance and thickness of artery and vein walls were measured in the thickest retinal vessels, which can be found around the optic nerve head (ONH). Measurements were collected from three healthy subjects. Retardation analysis and flow measurements were performed to locate vessel walls and to understand the polarization properties of blood vessel walls over time. Since the subject we imaged were in healthy condition, further studies on patients with cardiovascular disease are required to understand the impacts of diseases on blood vessel walls and the fibrous structures of the connective tissue.
2. Methods
2.1 Experimental setup
The PS-OCT system design was based on earlier designs [5,15]. The main difference between this design and earlier designs was the implementation of a 2.4-mm imaging aperture. Conventional OCT systems operate with a 1.2 mm beam diameter, the largest numerical aperture permitting aberration correction with a Badal optometer, which corrects for defocus [18]. The 2.4-mm beam size that was applied in this system provides an almost two times improved lateral resolution, a higher SNR and decreased speckle size [19]. The axial and lateral resolutions were approximately 6 μm and 12.9 μm, respectively.The axial resolution was calculated based on the central wavelength and optical bandwidth of the light source. The lateral resolution is based on the 2.4-mm 1/e2 beam width entering the eye, the estimated focal length of the eye (geometrical 22.3 mm, equivalent of 16.6 mm in air) and the center wavelength of the light source. Defocus was corrected by shifting the ophthalmic lens parallel to the beam path. Since astigmatism can affect image quality for beam widths over 1.2 mm [20], astigmatism was corrected with a trial lens.
2.2 Overview of scanning parameters
The field of view and vertical density (number of B-scans) were altered in four different configurations. Table 1 summarizes the implemented configurations. A larger number of B-scans per volume (utilizing different configurations) leads to (1) smaller time intervals between B-scans, (2) a reduction in speckle noise by averaging multiple B-scans obtained from the same location, and (3) a reduction of motion artifacts between adjacent frames introduced by involuntary eye motion. A larger number of B-scans helps when tracing a blood vessel in time and to visualize a blood vessel wall’s expansion and contraction due to blood flow pulsatility. For instance, in Fig. 3, the en-face image has 500 B-scans (using 3rd configuration) and hence, when tracing the blood vessel, 30 B-scans were assessed. However, if we had used 100 B-scans (1st configuration) instead, it was only possible to assess six B-scans in a similar time frame, which could result in loss of detail and information.
Table 1. Settings for different experiments to create en-face intensity, retardation and flow images.
The line scan camera used in the PS-OCT setup was limited to 25 kHz and owing to the 4 s measurement time of imaging, deemed a comfortable imaging time for subjects. The number of A-lines per B-scan was decreased as the numbers of B-scans increased. The smallest field of view (4.5 mm by 450 µm) was larger than the biggest blood vessel (∼ 200 µm) in the retina.
2.3 Subject recruitment and subject comfort
Three healthy subjects were recruited from the staff of the University of Western Australia. Informed consent was obtained prior to the experiments. Measurements were approved by the Human Research Ethics committee of the University of Western Australia and adhered to the tenets of the Declaration of Helsinki. In normal daylight the size of the human pupil varies from 2 to 4 mm and doubles in a dark environment [21]. Since our measurements were performed with a sufficiently small beam size of 2.4 mm in a dark environment, dilation of the pupils was not necessary. A chin and forehead rest were provided for the subjects to minimize motion artifacts. The eye that was imaged was provided with a fixation target to reduce unintentional eye motion.
2.4 Statistical analyses
Arterial and venous wall thickness, birefringence and overall retardation were determined and reported as averaged values of four measurements. For the inter-observer measurent, Pearson’s p-value, intraclass correlation coefficient (ICC) and Bland-Altman plots were used to compare the quantitative data. The statistical tests were performed using custom-written MATLAB (2019b, Math Works, Massachusetts, USA) scripts. Statically, a p-value of 0.05 was deemed significant. ICC ≥ 0.85 was considered excellent, 0.7 ≤ ICC ≤ 0.85 was classified as good, 0.5 ≤ ICC ≤ 0.7 was indicated as moderate and lower values were categorized as poor. The model to estimate ICC and their 95% confident intervals (CIs) were chosen based on the criteria provided elsewhere [22]. This model is based on a mean of two measurements, consistency, and a 2-way mixed-effects model. In Bland-Altman plots the difference between two independent examiners measuring the vessel wall thickness, double-pass phase retardation per unit depth (DPPR/UD) and overall retardation were plotted against the mean of the two observations. Limit of agreement (LOA) and corresponding percentage were calculated based on ±1.96 × standard deviation of the differences and (LOA/mean of the data) × 100, respectively [23]. LOAs smaller than 30% were considered reliable. Data points with larger LOA values were also displayed in the plots to reflect unreliable data.
2.5 Retardation measurements
Since diattenuation is considered to be insignificant in biological tissue and the angular deviation of Stokes vectors over a certain depth depends on tissue birefringence and noise, a Stokes-vector-based analysis was used to retrieve retardation values. In short, the collected spectra of the spectrometer were linearly mapped to k-space and compensated for chromatic dispersion [24–26]. After calculating the Stokes parameters I, Q, U and V for each pixel, pixels were re-aligned with respect to the retinal surface. This surface was located with an intensity threshold and used as the reference for retardation analysis. The Stokes parameters were then averaged with differently sized kernels (empirically determined) based on the image size (number of A-lines in a B-scan) listed in Table 2 to reduce the effects of speckle noise on the retardation [15]. Next, the Stokes vectors for each pixel were determined at the tissue surface and compared to those at a depth z to make the analysis independent of the incident polarization state [27], giving more weight to the largest angle measurement and the measurement with the highest SNR [28]. After obtaining a phase retardation image, the DPPR/UD, proportional to the blood vessel wall birefringence was obtained by fitting a first-order least squares linear fit [29]. In comparison, procedures such as comparing the retardation at the top and the bottom of the blood vessel wall are more sensitive to spurious data points. Fitting DPPR values made measurements less dependent on an accurate thickness measurement, which can be difficult due to the small size of the region of interest (the ∼15 μm thickness of a vessel wall) compared with the axial resolution (6 μm).
Table 2. Empirically derived moving average boxes used to generate retardation images in different applied configurations.
2.6 Determining the edge of a blood vessel wall
In the intensity B-scan, blood vessels are associated with a region of high intensity at the top of the RNFL followed in depth with an area with a lower intensity due to the higher forward scattering of moving blood cells [30]. Blood flow analysis can also help to determine the location where the blood vessel wall and the blood flow inside the vessel meet and relies on a color-Doppler analysis [31,32]. Figure 1 shows steps taken toward determining the edge of the blood vessel wall. In step (6), three criteria were empirically used to manually determine the blood vessel wall-lumen interface. Starting at the vitreous-retina interface, the cumulative DPPR increases linearly with depth in the birefringent vessel wall until it levels off at the vessel wall-lumen interface. The stabilization of cumulative DPPR with depth in the non-birefringent lumen is important, as the many pixels inside the lumen validate the magnitude of the cumulative DPPR that is induced by a few data points in the birefringent vessel wall. In earlier work, we used a similar approach to determine the thickness, DPPR and DPPR/UD of thick and thin retinal nerve fiber layer [33]. The intensity drops below the vessel wall-lumen interface (towards the photoreceptors) as forward scattering in the blood flow attenuates the intensity, the second criterion that helps to determine the vessel wall-lumen interface. At the same location, the Doppler signal changes due to flow, the third criterion. Of these three criteria, at least two were met to determine the edge of the wall. Note that the intensity inside the lumen is almost constant with slight alterations due to noise. The retardation measurement in tissue below the lumen, such as the lower vessel wall, is not reliable due to polarization scrambling in the blood flow and a low SNR [32].
Fig. 1. Six steps taken to determine the edge of the blood vessel wall and to retrieve the vessel wall polarization properties. In step (1) the proper B-scan in an intensity en-face image (obtained from healthy subject 3’s retina) was identified and in step (2) the logarithmic intensity (upper) and flow (lower) B-scans were generated. In the upper panel, blood vessels were associated with a region of high intensity at the top of the RNFL followed in depth with an area with a lower intensity. In the bottom panel, the color-Doppler image shows the blood flow. Depending on the direction of the blood flow, pixels are color-coded in white or red. Red arrows in steps (1 & 2) show the blood vessels. In step (3), the intensity and flow cross-sections were realigned with respect to the retinal surface. In step (4) the blood vessel boundaries (vertical red dotted lines) were determined with help of the intensity and flow images. In step (5) the area between the vertical lines was isolated and rotated by 90°. The X- and Y-axes now represent the depth and A-line numbers, respectively. Step (6) shows the intensity, flow and retardation plots. These plots are the result of data averaging over all A-lines within the area between the two vertical dashed red lines in step (4). Intensity, flow and cumulative retardance data were then used to estimate the edge of the blood vessel wall. A decrease in the intensity attributed to the blood flow inside the vessel (step (6), black line) was observed inside the wall. The cumulative retardation increases linearly with depth in a vessel wall, after which it levels off (step (6), red line). Decreased blood flow values (step (6), green line) can also help to determine where the blood flow starts. The retardation plot is fit with a least-squares linear fit. Finally, the cumulative retardation value at the wall’s edge is divided by the thickness of the blood vessel wall to calculate the DPPR/UD. A blue dashed arrow in step (6) shows the edge of the blood vessel wall. The thickness of the marked vessel wall is 12 µm and its DPPR/UD is 1.01°/μm. Movie file (Visualization 1): step (2) shows the flow movie. The artery and vein branching out vertically from the ONH have the same blood flow direction, likely due to the angle between the blood flow direction and the OCT incident beam, governing the measured flow in the axial direction.
3. Results
The results section is divided into several parts, each with its own objective. We will start by explaining the accuracy and precision of our PS-OCT method, then we evaluate the impacts of averaging on determining the blood vessel wall. Next, we quantify tissue properties along the same vessel and visually show the variation of thickness and DPPR/UD along a vessel wall. After looking for differences in artery and vein wall properties near the ONH, the DPPR/UD of the blood vessel wall and adjacent RNFL tissue are compared to investigate whether measurements were performed inside the vessel wall or in covering nerve fiber tissue. Afterwards, we determine the influence of involuntary eye motion on the method by tracing three artery/vein pairs. This was followed by an experiment in which the same pairs were evaluated by a second examiner to determine the inter-observer agreement (IOA). Finally, the repeatability of the proposed method was determined for a subject during four visits.
3.1 Accuracy and precision of the PS-OCT system
Accuracy and precision affect the thickness measurement of the blood vessel walls. We therefore quantified these two properties by imaging a plane mirror with PS-OCT, set at a distance of 575 μm from 0 μm optical path length difference. The accuracy of the measurement can be defined as the ratio between the difference between the actual distance measured by a micrometer and the mean over the measured value, while the precision can be determined by the ratio of standard deviation and mean [34]. The standard deviation and mean were determined from 1000 A-lines in a single B-scan. The distance varied between 572.60 μm and 576.68 μm with a mean and standard deviation of 573.78 μm and 0.82 μm, respectively. The accuracy was therefore 1.22 μm (0.21%) and the precision was 0.82 μm (0.14%). The same experiment was repeated, with a distance range of 572.56 μm and 576.87 μm, a mean of 574.11 μm and standard deviation of 0.94 μm. The accuracy and precision for the second experiment were 0.15% and 0.16%, respectively. The accuracy was affected by our calibrated pixel to μm conversion in the software, which with a value of 575 μm / 138 pixels = 4.17 μm/pixel (3.02 μm/pixel in tissue, vs. 3.05 μm/pixel for our measurements) would have resulted in a higher accuracy. The precision is systematic and stems from shot noise. Our measurements are affected not only by shot noise but by speckle noise as well and involve fewer A-lines for speckle-noise and shot-noise averaging. The precision in the vessel wall measurements is therefore lower.
3.2 Measurements in the vertical and horizontal blood vessel wall
To evaluate how the region of interest affects the DPPR/UD and thickness measurements, DPPR plots in the vertical and horizontal vessel walls were obtained (Fig. 2). The vertical wall covers a longer distance than the horizontal wall, and it is encouraging that its birefringence is very similar to the measurements in the horizontal vessel wall, which was obtained with fewer pixels, further strengthening the reliability of these birefringence measurements and the contrast in birefringence between the wall and the RNFL tissue.
Fig. 2. (A) Flow cross-section realigned with respect to the retinal surface indicating the blood vessel boundaries (vertical red dotted lines) of subject 3. (B) Flow cross-section rotated by 90° showing the different regions. (C) Retardation plots of A-lines 1-7 and 42-49 traversing the vertical vessel wall (B). (D) Retardation plots of the four regions shown in panel (B). A kink was observed in the retardation plots isolated between A-lines 1-7 and 42-49. Near the top of the retina, lower DPPR/UD values are retrieved with a least square linear fit followed by a slope with relatively high DPPR/UD, suggesting that the highly birefringent blood vessel wall is covered by RNFL with a lower birefringence. In contrast, a kink was not observed in the horizontal vessel wall in A-lines 21-28, indicating that this area was not covered by RNFL tissue. The flow image size in (A) is 350 µm by 3.0 mm.
3.3 Vessel wall properties along a vessel
A tracing procedure was followed to determine how much blood vessel wall properties change along a section of a vessel (Fig. 3). While the thickness remains mostly constant in this section of the vessel (Fig. 3(A)), its DPPR/UD varies by 15% (Fig. 3(B)).
Fig. 3. (A) En-face image of the blood vessels located near the optic nerve of the right eye of subject 2 with the 3rd scanning configuration. The en-face image measures 4.5 mm by 0.9 mm. B-scans in the area located between two horizontal green lines panel including B-scans 336 to 365 were used to calculate DPPR/UD and thickness of the indicated blood vessel (white arrow) wall. Panel (B) shows the variation in DPPR/UD and the thickness of the blood vessel wall as a function of location/B-scan number. Only cross-sectional images with blood vessels and a high SNR were included in the measurements. While the thickness remains mostly constant in this section of the vessel, its DPPR/UD varies by 15%.
3.4 Artery and vein classification
Artery and vein walls are differently affected by retinal, cerebral and systemic disorders [4]. It is therefore helpful to classify blood vessels into arteries and veins. There is a difference based on the appearance of the blood vessels near the ONH [35]. The arteries are thinner due to a narrow lumen and have a lower reflected intensity. Arteries also look less bright in B-scans and tend to be curlier compared to veins. Due to a larger width, more of the incoming OCT light is blocked by the veins, creating larger shadows below the vessels in B-scans. Veins look brighter in the en-face images. Figure 4(A) shows the retinal fundus image centered at the optic nerve head of the right eye of subject 3. Based on reflected intensity and vessel thickness, the marked blood vessel pair is classified as artery and vein.
Fig. 4. (A) Retinal fundus and (B) en-face intensity image of the right eye of subject 3. Both images measure 3 mm by 3 mm. Whilst the visual differences between arteries and veins in the en-face OCT intensity images help to differentiate arteries from veins, the DPPR/UD and thickness of the blood vessel wall together with the introduced overall retardation can provide additional contrast to discriminate between the two blood vessel types. In this example, the artery and vein classification in the fundus image (A) was consistent with the artery/vein classification determined based on PS-OCT results (B). In the PS-OCT results, arteries consistently have higher DPPR/UD and overall retardation values.
The vessel walls of arteries are believed to be thickest [4], due to a thicker tunica media, the second layer in an artery or vein wall. This tunica media is made of smooth muscle cells and elastic fibers that enable vessels to stretch, causing a change in artery thickness during a heartbeat, unlike the behavior of veins, which experience a much smaller thickness change. Since the arteries are exposed to higher pressures, a thicker muscle layer provides higher flexibility to accommodate the changes in volume associated with these pressure changes [36,37]. It would be unwise to use the wall thickness as the sole parameter to discriminate between arteries and veins. Many factors are involved in the wall thickness of an artery or vein such as the location of the blood vessels in the retina and their overall size. Systemic cardiovascular and eye disorders can also affect the thickness of a blood vessel wall in the retina. Both the induced retardance and vessel wall DPPR/UD may be properties that can be used alongside the wall thickness to reliably differentiate arteries from veins. For instance, in Fig. 4(B) (en-face intensity image; subject 3), the classification was done based on the higher values of DPPR/UD (0.51°/µm vs. 0.48°/µm) and overall retardation (9° vs. 6°) of the retinal artery over the retinal vein, respectively. Looking at the reflectivity and morphology in the fundus image, this classification is correct.
More artery/vein pairs were classified based on their DPPR/UD, overall retardation and thickness values. Orange lines in Fig. 5 (en-face intensity images taken from various subjects) mark the B-scans with blood vessel pair classification. Information collected from these pairs is provided in Fig. 6.
Fig. 5. En-face intensity images taken from (A) the left and (B) the right eye of subject 1, and the right eyes of (C) subjects 2 and (D) 3. The orange lines in (B), (C) and (D) are the B-scans for artery/vein classification in Fig. 6. Red and blue arrows show the arteries and veins, respectively. Red solid lines show the B-scans chosen to compare birefringence of blood vessel walls (indicated here with red arrows and numbered accordingly) and adjacent RNFL tissue presented in Fig. 8. Dark areas in the figures depict the locations with higher intensities. In all images, the X-axis represents the A-line number; the Y-axis represents the B-scan number.
Fig. 6. Extracted thickness (blue/cyan), DPPR/UD (red) and accumulated retardation (grey) of three artery-vein pairs. Both the artery and vein were found in the same B-scan, near each other with similar widths (orange lines in Fig. 5). The represented data are the average of 4 adjacent B-scans. Error bars indicate the standard errors of these four measurements. The DPPR/UD of the arteries are typically higher than those of veins. In case of overlapping wall DPPR/UDs (subject 3), the accumulated retardation (grey) helps to discriminate between the artery and vein.
According to Fig. 6, the DPPR/UD and overall retardation of the artery walls were consistently larger than those of the vein walls (red, grey). For subject 1, the value of DPPR/UD for the artery and vein wall were 0.78°/µm and 0.52°/µm, respectively. The overall retardation was also higher for the artery compared to the vein (12° and 9°, respectively). Subject 2’s artery had a higher DPPR/UD value of 0.75°/µm compared to that of vein, which was 0.57°/µm. The walls of artery/vein introduced overall retardation of 11° and 9°. The same trend was observed for the third artery/vein pair in subject 3’s retina. As discussed previously in Fig. 4(B), DPPR/UD and overall retardation of the artery wall were higher compared to those of the vein.
3.5 Comparing DPPR/UD of blood vessel walls and adjacent RNFL tissue
We looked at a vein in the left eye of subject 1 (Fig. 7(A), red arrow) and extracted cross-sectional intensity and flow images (Fig. 7(B)) of the indicated B-scan that is represented by the horizontal red line in Fig. 7(A). After choosing the boundaries of the vein (Fig. 7(B), the area between vertical red dashed lines), retardation and intensity plots as a function of depth were retrieved (Fig. 7(C)). Starting from the top of the RNFL (pixel zero), the retardation linearly increased as a function of depth in the area that we believe is inside the blood vessel wall and plateaued at depths larger than ∼15 μm, an indication that the vessel wall does not extend beyond this point. The retardation measurement is less reliable in the blood vessel or below the vessel, due to blood flow polarization scrambling. The normalized intensity plot seems less useful here, because the log intensity is noisy, presumably due to the blood flow in the vessel. Starting from 0°, the retardation increased to approximately 12° at a thickness of 15 μm. A linear least-squares fit of the data resulted in a DPPR/UD of the blood vessel wall of 0.77°/µm. Although the intensity plot is noisy, a drop in intensity (determined empirically) indicates the vessel wall-lumen interface. These layers can be seen in Fig. 7(C) in detail.
Fig. 7. (A) En-face intensity image of the ONH in the left eye of subject 1 using the 1st scanning configuration. The image measures 4.5 mm by 4.5 mm. Panel (B) shows the logarithmic intensity and flow images of the B-scan (image size is 365 µm in depth by 4.5 mm laterally) indicated by the red arrow in (A). Black arrows in panel (B) show different regions of interest. Panels (C) and (D) show the retardation and intensity plots averaged over all A-lines of the blood vessel (constrained between the two dashed red lines, between A-lines 412 to 451) and an adjacent area of RNFL, marked by the two vertical yellow lines. A least-squares linear fit was applied to the retardation plot to determine the DPPR/UD and the thickness of the blood vessel wall and RNFL. The retardation plot in panel (C) shows a linear increase with depth inside the blood vessel wall up to approximately 15 µm. After this point the normalized log intensity experiences a drop and retardation reaches a plateau, indicating the edge of the wall/lumen interface. From the intensity plot we can estimate that the blood vessel is approximately 80 µm thick. The DPPR/UD of the blood vessel wall was 0.77°/µm. The DPPR/UD of the RNFL on the other hand, determined in panel (D), was 0.43°/µm, which is smaller than that of the blood vessel. GCL/IPL complex: Ganglion cell layer/inner plexiform layer complex.
To better compare the differences between tissue properties of the RNFL and the blood vessel wall, the DPPR/UD of the RNFL to the left of the blood vessel (area between yellow dotted lines in Fig. 7(B)) was determined. Retardation and intensity plots of the RNFL tissue, Fig. 7(D), followed the same pattern which we saw previously in blood vessel walls. Retardation increased linearly inside the RNFL and levelled off because of the non-birefringent nature of the ganglion cell layer and inner plexiform layer (IPL). The DPPR/UD of the RNFL tissue was 0.43°/µm. The vein’s wall thickness was 15 µm and the induced retardation was 12°. In comparison, the accumulated retardation for the same thickness of RNFL was only 6°, underscoring the difference in birefringence of the two types of tissue, and confirming that the measurements in the vessel wall are indeed obtained from vessel wall tissue, not from RNFL tissue that covers the blood vessel wall.
The difference in DPPR/UD of the blood vessel walls and adjacent RNFL tissue were further investigated in a larger population. These arteries and veins are identified in Fig. 5 with red and blue arrows. Solid horizontal red lines mark the B-scans that were used to calculate thickness and DPPR/UD. Figure 8 summarizes the measurements obtained from arteries, veins, and the adjacent RNFL tissue. In this figure, arteries are shown in red and veins are shown in blue. The reported values are the result of averaging data over four adjacent B-scans. The DPPR/UD of the blood vessels walls were consistently larger than the DPPR/UD of the RNFL tissue. For instance, blood vessel #1 (an artery located in the retina of the left eye of subject has a higher DPPR/UD (0.61°/µm) than neighboring RNFL tissue, which was 0.19°/µm.
Fig. 8. DPPR/UD, thickness, and accumulated retardation of artery walls, vein walls and adjacent RNFL tissue. The numbered blood vessels are displayed in Fig. 5. Blood vessels #1 to #12 stand for a combination of the arteries, veins and their adjacent RNFL tissue of the left and right eyes of subject 1, and the right eye of subjects 2 and 3. Data were averaged over 4 adjacent B-scans and error bars represent the standard error. The DPPR/UD of the walls of the blood vessels were larger than the DPPR/UD of the RNFL tissue. The induced double pass retardance is consistently larger in the vessel wall, confirming that measurements were obtained from vessel wall tissue and not from overlying nerve fiber tissue.
As a note, it is not possible to draw a conclusion about the artery-vein classification in these panels based on the DPPR/UDs and overall retardations (DPPR/UD × thickness), since these arteries and veins were selected randomly from different subjects and from different location of the retina. Within the same subject veins sometimes have a higher DPPR/UD, overall retardation and thickness value compared to arteries. Blood vessels #1 to #3 are good examples. While the wall of the artery (BV#1) only introduced 8° of retardation (0.61°/μm × 14 μm), the retardation for the vein walls (BV #2 and #3) were 10° and 9°, respectively. Consequently, classification with PS-OCT should be used to classify neighboring artery/vein pairs.
3.6 Motion artefact-induced errors
Repeated measurements were performed on the same artery and vein (right eye of subject 1) to determine how much measurements vary when they are obtained from the same location, for instance due to involuntary eye motion. In these measurements we used a repeated B-scan. The intensity and retardation en-face images of the repeated B-scan can be seen in Fig. 9 (B-scan #1, #2 and #3). The vertical arrows in the figure indicate the assessed artery and vein pairs over time. Detailed data are available in Fig. 10. B-scan #1 has the following values for DPPR/UD, thickness and the overall retardation of the artery’s wall: (0.57 ± 0.04)°/µm, (17 ± 1) µm and (9.5 ± 0.7)°, respectively. For the vein, these values were: (0.49 ± 0.05)°/µm, (17.0 ± 1) µm and (8.2 ± 0.6)°, respectively. The other two artery/vein sets followed similar trends meaning that artery walls were thinner compared to vein walls, while the DPPR/UD was higher in artery walls. This is consistent with results we found in the previous section. The processed values in Fig. 10(A) are the average of acquired B-scans and the error bars indicate the standard errors. The variations of DPPR/UD and thickness and their averages can be seen in different B-scans in Fig. 10(B) to (D).
Fig. 9. En-face (A) intensity and cumulative retardation images (B) taken from the right eye of subject 1. Red solid lines show the location of repeated B-scan. Panels (C) and (D) show the repeated B-scans (#1) over time; (E) and (F) show the repeated B-scans (#2) and (G) and (H) show the repeated B-scans (#3). Red and blue arrows mark the arteries and veins, respectively. For all panels, the X-axis represents the A-line number and Y-axis the B-scan number, in this case it represents time as well. Image sizes (A) and (B) are 4.5 mm by 4.5 mm.
Fig. 10. (A) Averaged DPPR/UD, thickness and overall retardation of the walls extracted from sets of artery/veins in B-scans #1 to #3 indicated in Fig. 9. In these measurements, a repeated B-scan was applied to scan the same location. In these sets arteries typically have a smaller wall thickness compared to veins, while the DPPR/UD is higher in arteries. Subsequently, overall retardation of the blood vessel wall is higher in arteries than in veins for a similar thickness. The processed values in (A) are the average of all acquired B-scans and the error bars indicate the average standard errors. The panels B, C, and D show variations in DPPR/UD and wall thickness over time in artery/vein sets #1 (B), set #2 (C) and set #3 (D). Dots show the actual measured DPPR/UD and thickness and dotted lines represent the averaged values reported in (A).
The standard deviations are about 14% of the average DPPR/UD and thickness values, indicating that involuntary motion artefacts are not to be ignored.
3.7 Inter-observer agreement
The inter-observer agreement was determined to investigate how two examiners (H.A. and F.N.) analyze the same data. This is important since different examiners can have different interpretations of the blood vessel wall boundaries and consequently end up with different values. This inconsistency can affect the method in a clinical setting. According to the data collected by each examiner (listed in Table 3), the values are almost the same when measuring DPPR/UD. A small inconsistency can be seen in thickness and overall retardation. This discrepancy can be retraced to discrete values of wall thickness in our PS-OCT thickness measurement. The average values and standard errors determined by two examiners overlapped well.
Three different statistical analyses were also performed to evaluate consistency of the data acquired by two examiners. Table 4 provides a summary of the ICC, Pearson and Bland-Altman analysis. Based on ICC values, the DPPR/UD of the blood vessel walls revealed good to excellent (2 cases) and excellent (4 cases) consistency between the two examiners. For the thickness, the ICCs range from 0.391 to 0.595 which are classified poor and moderate, which is the result of discrete wall thickness values. Likewise, the overall retardation, which is DPPR/UD × thickness, showed moderate ICC values (0.573 to 0.721). Based on the ICC values, the DPPR/UD measurements are independent of the examiner.
Pearson p-value for DPPR/UD, thickness and overall retardation (mostly) were smaller than <0.0001 which indicates that the two examiners agree well. The only p-value which is smaller than <0.01 is related to the artery measurement in B-scan #2. Although this p-value is significant, the mean and standard deviation of results obtained by each examiner were almost the same. For example, H.A. measured an artery thickness of 15 ± 1 µm, while F.N. determined a thickness of 16 ± 1 µm, indicating an overlap between the two examiners.
Figure 11 shows the Bland-Altman plots for the artery/vein sets of B-scan #1, 2 and 3. Based on this Fig. and data represented in Table 4, the mean differences (bias) for the DPPR/UD are between -0.014 and 0.06. The LOA ranged between 12 to 17% (excellent). For the thickness and accumulated retardation, LOAs were between 26 to 29% (excellent) and 27 to 33% (good), respectively. Data points outside the LOA range were less than 6%. Bland-Altman analysis indicates that there is no significant discrepancy in either of the measurements.
Fig. 11. Agreement between the two examiners using Bland–Altman plots. These plots show the DPPR/UD of the same blood vessel over time, in artery/vein sets shown in Fig. 9. The vertical axis shows the difference between DPPR/UDs, determined by the two examiners. The horizontal axis displays the mean of the DPPR/UD measurements determined by the two examiners. The black lines represent the bias, which is the average of the differences. The dotted lines are LOAs (bias ± 1.96×standard deviation of the differences). It can be seen from the biases in the DPPR/UD measurements that there is good agreement between the two examiners. By comparing LOAs and the mean of the measurements for each blood vessel, LOAs are small, again indicating good agreement. Superscripts indicate whether the measurements were performed by examiner 1 or 2.
In summary, although the ICC analysis of thickness and retardation indicates that the two examiners do not always agree, the p-values and Bland-Altman plots reflect a strong agreement, demonstrating that the method is independent of the examiner.
3.8 Repeatability
The repeatability was determined by analyzing an artery/vein set of the right eye of subject 1, imaged during four visits. Figure 12 shows the averaged thickness, DPPR/UD and overall retardation of four adjacent B-scans obtained during those four visits. Error values represent the average standard errors of these four B-scans. DPPR/UD for the artery ranged between 0.53°/µm and 0.58°/µm and the average standard error was ±0.02°/μm. The values for the vein were 0.40 to 0.56°/µm with a standard error of ±0.02°/μm. Standard errors are ≤5% of the DPPR/UDs. The thickness of the walls varied between 12 μm and 14 µm for the artery and 14 µm for the vein. Considering that the pixel size is 3.04 µm, the standard errors for wall thickness are smaller than the pixel size. The same behavior was observed for the overall retardation: overall artery retardations ranged from 6.7° to 7.4° with a standard error of ±0.5°. Those of veins ranged between 5.5° and 8.2° with an average standard error of ±0.4°. At ≤5%, the repeatability of the DPPR/UD measurement over four visits is better than the ≤7% standard error obtained from the same tissue in a repeated B-scan (Fig. 10), demonstrating that the same tissue can be probed reliably during subsequent visits.
Fig. 12. Repeatability measurements showing averaged values for thickness, DPPR/UD and overall retardation obtained over four adjacent B-scans during four visits. Data was collected from the right eye of Subject 1. The error bars represent the standard error. For DPPR/UD, the standard error is less than 5%. The standard error for the thickness is <6% which is smaller than the pixel size of 3.04 µm. The standard error was again less than 6% for the accumulated retardation. The averaged retardation for the artery and vein were 7.2° and 6.8°, respectively.
4. Discussion
Accuracy and precision of less than a micrometer performed on a plane mirror showed that the thickness measurements can be carried out with a roughly 1 μm precision.
Measurements were performed on arteries and veins near the optic nerve head of healthy human subjects to determine the vessel’s polarization properties. While the wall thickness did not provide enough contrast for artery/vein classification, the additional DPPR/UD and overall cumulative double pass retardance provideda clear distinction between the two, when an artery/vein pair is directly compared. The DPPR/UD of a single vessel is not sufficient to classify it as an artery or vein. Within pairs, higher DPPR/UD and retardance values were found in artery walls, presumably due to the presence of smooth muscle cell tissue. In Table 5 the measurements are compared to literature values.
Table 5. Overview of the literature: reported blood vessel wall thickness a
So far, in-vivo blood vessel wall thickness studies were based on SD-OCT intensity data [4,40], OCT-angiography (OCTA) data [41], and adaptive optics scanning laser ophthalmoscopy (AOSLO) data [38], or scanning laser Doppler flowmetry (SLDF) data [41]. The results from various studies are not always in agreement with each other. For instance, Chui et al. [38] performed AOSLO measurements on eight different subjects and showed that mural cells (principal cells regulating the stability and homeostasis of vasculature) in the retinal blood vessel walls cause retinal arteries to have more than 10 µm thicker artery walls. Other studies concluded that although the arteries have thicker walls, the difference is less than 3 µm [4,39,40]. Rim et al. [4] showed that the inner wall thickness (toward the vitreous) of the arteries and veins have thicknesses of 23.9 µm and 20.7 µm, respectively. They claimed that by using SD-OCT intensity-assisted images, they were able to more objectively separate the blood vessel wall from surrounding tissue and blood flow. Muraoka et al. on the other hand found vessel thickness values of 14.0 µm and 11.7 µm for retinal arteries and veins, respectively, with very high reproducibility [39]. Moreover, color fundus photographs showed that the wall thickness of the retinal arteries and veins ranged between 14.2 µm and 12.2 µm, respectively [40]. These studies were done on similarly sized arteries and veins near the ONH of young healthy subjects. The vein walls were thicker than artery walls in measurements with SLDF [41]. Histology on the other hand showed that the wall thickness of the retinal arteries is ∼18 µm and retinal veins is ∼14 µm [4]. Although histology is considered to be the gold standard, it is not an in-vivo measurement and cannot be used to monitor the onset and progression of different diseases. It is believed that the discrepancy in the various studies regarding the retinal blood vessel walls thickness may be caused by 1) a lack of contrast in OCT intensity images, 2) incorrect determination of blood vessel wall boundaries, 3) inter-subject variation in eye properties, 4) errors during data processing and 5) blood flow changes during the cardiac cycle, affecting the wall thickness estimation.
In our study, artery walls on average have a DPPR/UD of 0.72°/μm, while the vein wall’s DPPR/UD is on average 0.58°/µm. The overall retardation of the artery wall is also higher than the retardation induced by the vein wall (on average: 10.8° compared to 8.7°). The average thickness of the artery and vein walls adjacent to the ONH are ∼15 μm for both, which is different from the previous studies, emphasizing the importance of retardation and DPPR/UD measurements for artery/vein classification, in particular for extreme cases where the veins have thicker walls. For example, in Fig. 6, the artery/vein wall thicknesses of subject 1 are (16 ± 1) μm and (18 ± 1) µm, respectively, and those of subject 2 are (15 ± 1) μm and (17 ± 1) µm, respectively, which is in contradiction with the assumption that arteries have thicker walls. However, retardation for these cases indicated that the arteries have induced a higher retardation than the veins. On average, the retardation values of the first set is (12 ± 1)° for the artery and (10 ± 1)° for the vein and the retardation of the second artery/vein set are (11 ± 1)° and (6 ± 1)°, respectively, which matches our hypothesis. The higher DPPR/UD values in arteries were expected since the arteries have thicker layers of muscle-composed tunica media [36,37].
We compared the DPPR/UD of the retinal blood vessel walls to the DPPR/UD of the RNFL. A previous study reported that the birefringence of the RNFL was 1.2 × 10−4 temporal and 4.1 × 10−4 superior and inferior to the optic nerve head at 840 nm [33]. For better comparison, we converted the DPPR/UD values from this study to dimensionless birefringence at 840 nm. The RNFL of the left eye of the subject 1’s eye is 2.25 × 10−4 temporal, 3.5 × 10−4 superior and 4.4 × 10−4 inferior to the ONH. Hence, the birefringence of the RNFL in the current study is in agreement with values reported in the literature. For comparison, the birefringence of the blood vessels adjacent to the RNFL are: 7.11 × 10−4, 6.94 × 10−4 and 7.50 × 10−4, respectively. This demonstrates that the vessel walls are not covered by RNFL tissue, and that their birefringence is considerably larger than the RNFL birefringence. If a thin layer of RNFL covers the blood vessels, the DPPR induced by this layer over its thickness would be negligible, as the DPPR induced by the much thicker and more birefringent vessel wall would dominate the least squares fit to determine the DPPR/UD.
Furthermore, the smooth muscle tissue in larger cardiovascular vessels has a birefringence of around 6 × 10−4 [42]. This is also in good agreement with our findings since we believe that the large values of birefringence in the blood vessel walls comes from the smooth muscle cells in the tunica media.
We evaluated the repeatability of our retardation measurements on retinal blood vessel walls by measuring the retardation multiple times at the same location. According to Fig. 10(B) to (D), the DPPR/UD varied by approximately 7% (based on the standard error) demonstrating a good repeatability. For the thickness measurement, the variation was also 7% to 8%. Relying on discrete thickness values with a multiple of 3.04 μm, and an average thickness of vein and artery walls of (17 ± 1) µm and (17 ± 1) µm, respectively (first and second panel, Fig. 10(B)). Being off by one pixel causes a difference of ∼17%, demonstrating that the thickness estimation is at a sub-pixel level. Furthermore, wall thickness variations are in good agreement with values reported in the literature [4,39, 40]. The reason behind the relatively high level of discretization is the limited axial resolution of our PS-OCT system, due to the 50 nm full width at half maximum optical bandwidth of the broadband source. One way to decrease the effects of discretization of the thickness values is to use an ultrahigh resolution PS-OCT system [43].
If we consider our PS-OCT retinal blood vessel wall measurements for clinical use, there are several limitations. First, although the system’s axial resolution of ∼6 µm is sufficient to discriminate between artery and vessel pairs walls based on their polarization properties, an accurate thickness measurement requires ultrahigh PS-OCT. Second, the examiner affects the results, but p-values, ICCs and Bland-Altman analysis demonstrate a high degree of agreement between the two examiners. To further improve the consistency of the measurements, artificial intelligence could replace the examiner, which could lead to a reduction of errors.Third, in this study only normal eyes of young healthy subjects were imaged, typically with high SNR. Lower SNR causes the retardation measurements to be unreliable [32]. Reduction in lens opacity or cataract, which are expected to occur in older patients, would result in less accurate retardation measurements, unless the sensitivity of the system can be increased, for instance with a larger aperture size [44].
This study has also demonstrated the feasibility of this non-invasive measurement as a practical tool for vessel wall assessment. Unlike the intensity-based SD-OCT wall thickness measurements that can only rely on intensity variations for wall thickness measurements, our method determines the wall thicknesses based on intensity, retardation and flow. The system requires ∼4 s to record a volumetric scan of the eye, making it practical for a clinical environment. The most important benefit of this technology is the extra contrast that PS-OCT offers: the ability to not only show the thickness, but also quantify the polarization properties which are proportional to the structural integrity of retinal blood vessel walls.
4.1 Future work
So far, measurements have been performed in healthy subjects only. To test whether the method is feasible for discrimination between healthy and diseased tissue, additional measurements will have to be obtained from patients with cardio-vascular disorders.
5. Conclusion
We presented a method based on PS-OCT to quantify the thickness, birefringence and cumulative phase retardation of retinal blood vessel walls. We demonstrated that our method reliably distinguishes between artery and vein pairs and provides more contrast than an intensity OCT measurement alone. By comparing the results of blood vessel walls with those obtained from adjacent RNFL tissue, it was demonstrated that the vessel walls near the optic nerve are not covered by RNFL tissue. The DPPR/UD, cumulative retardation and thickness measurements combined together are a robust tool to quantify blood vessel wall. Two examiners who analyzed the same data sets agreed well, as was evident from the statistical analysis of the data.
There are many grounds for future work, from studying the effects of cardiovascular diseases on retinal blood vessel wall integrity to utilizing ultrahigh resolution PS-OCT for improved thickness measurements. Machine learning will also speed up the analysis and diminishes the influence of human examiners.
Funding
Australian National Health and Medical Research Council APP1180854; National Research Foundation of Korea (2019H1D3A2A02101784); University of Western Australia.
Acknowledgments
H.A. and Q.W. acknowledge the financial support of the Australian Government International Research Training Program (RTP) Fee Offset and University Postgraduate Award (UPA) scholarships. M.J.H. would like to acknowledge the financial support of the University Postgraduate Award (UPA), SIRF scholarship and funding through the Ideas Grant from the National Australian Health and Medical Research (NHMRC) Council (APP1180854). S.V.J. acknowledges the financial support from the University Postgraduate Award (UPA) and Graduate Woman (WA) Research Scholarships. B.C. and C.J. acknowledge funding through the Brain Pool Program from the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (2019H1D3A2A02101784).
Disclosures
B.C. Massachusetts General Hospital (P). The other authors declare that there are no conflicts of interest related to this article.
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. A. Grosso, F. Veglio, M. Porta, F. M. Grignolo, and T. Y. Wong, “Hypertensive retinopathy revisited: some answers, more questions,” Br. J. Ophthalmol. 89(12), 1646–1654 (2005). [CrossRef]
2. R. A. Wimmer, A. Leopoldi, M. Aichinger, D. Kerjaschki, and J. M. Penninger, “Generation of blood vessel organoids from human pluripotent stem cells,” Nat. Protoc. 14(11), 3082–3100 (2019). [CrossRef]
3. S. Gulec, “Early diagnosis saves lives: focus on patients with hypertension,” Kidney Int. Suppl. 3(4), 332–334 (2013). [CrossRef]
4. T. H. Rim, Y. S. Choi, S. S. Kim, M. j. Kang, J. Oh, S. Park, and S. H. Byeon, “Retinal vessel structure measurement using spectral-domain optical coherence tomography,” Eye 30(1), 111–119 (2016). [CrossRef]
5. B. Cense, Q. Wang, S. Lee, L. Zhao, A. E. Elsner, C. K. Hitzenberger, and D. T. Miller, “Henle fiber layer phase retardation measured with polarization-sensitive optical coherence tomography,” Biomed. Opt. Express 4(11), 2296–2306 (2013). [CrossRef]
6. 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]
7. D. Huang, E. Swanson, C. Lin, J. Schuman, W. Stinson, W. Chang, M. Hee, T. Flotte, K. Gregory, and C. Puliafito, “Optical coherence tomography,” Science 254(5035), 1178–1181 (1991). [CrossRef]
8. J. F. de Boer, T. E. Milner, M. J. C. van Gemert, and J. S. Nelson, “Two-dimensional birefringence imaging in biological tissue by polarization-sensitive optical coherence tomography,” Opt. Lett. 22(12), 934–936 (1997). [CrossRef]
9. D. C. Adams, M. V. Szabari, D. Lagares, A. F. McCrossan, L. P. Hariri, A. M. Tager, and M. J. Suter, “Assessing the progression of systemic sclerosis by monitoring the tissue optic axis using PS-OCT,” Sci. Rep. 10(1), 2561 (2020). [CrossRef]
10. B. Cense, T. C. Chen, B. H. Park, M. C. Pierce, and J. F. de Boer, “In vivo depth-resolved birefringence measurements of the human retinal nerve fiber layer by polarization-sensitive optical coherence tomography,” Opt. Lett. 27(18), 1610–1612 (2002). [CrossRef]
11. J. F. de Boer, C. K. Hitzenberger, and Y. Yasuno, “Polarization sensitive optical coherence tomography - a review,” Biomed. Opt. Express 8(3), 1838–1873 (2017). [CrossRef]
12. E. Götzinger, M. Pircher, W. Geitzenauer, C. Ahlers, B. Baumann, S. Michels, U. Schmidt-Erfurth, and C. K. Hitzenberger, “Retinal pigment epithelium segmentation by polarization sensitive optical coherence tomography,” Opt. Express 16(21), 16410–16422 (2008). [CrossRef]
13. T. Torzicky, S. Marschall, M. Pircher, B. Baumann, M. Bonesi, S. Zotter, E. Gotzinger, W. Trasischker, T. Klein, W. Wieser, B. Biedermann, R. Huber, P. Andersen, and C. K. Hitzenberger, “Retinal polarization-sensitive optical coherence tomography at 1060 nm with 350 kHz A-scan rate using an Fourier domain mode locked laser,” J. Biomed. Opt. 18(2), 026008 (2013). [CrossRef]
14. J. Willemse, M. G. O. Grafe, F. D. Verbraak, and J. F. de Boer, “In Vivo 3D Determination of Peripapillary Scleral and Retinal Layer Architecture Using Polarization-Sensitive Optical Coherence Tomography,” Transl Vis Sci Techn 9 (2020).
15. B. Cense, D. T. Miller, B. J. King, T. Theelen, and A. E. Elsner, “Measuring polarization changes in the human outer retina with polarization-sensitive optical coherence tomography,” J. Biophotonics 11(5), e201700134 (2018). [CrossRef]
16. A. Weber, A. E. Elsner, M. Miura, S. Kompa, and M. C. Cheney, “Relationship between foveal birefringence and visual acuity in neovascular age-related macular degeneration,” Eye 21(3), 353–361 (2007). [CrossRef]
17. D. A. VanNasdale, A. E. Elsner, T. Hobbs, and S. A. Burns, “Foveal phase retardation changes associated with normal aging,” Vision Res. 51(21-22), 2263–2272 (2011). [CrossRef]
18. L. N. Thibos, X. Hong, A. Bradley, and X. Cheng, “Statistical variation of aberration structure and image quality in a normal population of healthy eyes,” J. Opt. Soc. Am. A 19(12), 2329–2348 (2002). [CrossRef]
19. M. Reddikumar, A. Tanabe, N. Hashimoto, and B. Cense, “Optical coherence tomography with a 2.8-mm beam diameter and sensorless defocus and astigmatism correction,” J. Biomed. Opt. 22, 026005 (2017). [CrossRef]
20. L. N. Thibos, X. Hong, A. Bradley, and X. Cheng, “Statistical variation of aberration structure and image quality in a normal population of healthy eyes,” J. Opt. Soc. Am. A 19(12), 2329–2348 (2002).
21. R. H. Spector, “The pupils,” in Clinical Methods: The History, Physical, and Laboratory Examinations. 3rd edition (Butterworths, 1990).
22. T. K. Koo and M. Y. Li, “A Guideline of Selecting and Reporting Intraclass Correlation Coefficients for Reliability Research,” J. Chiropr. Med. 15(2), 155–163 (2016). [CrossRef]
23. D. Giavarina, “Understanding Bland Altman analysis,” Biochem. Med. 25(2), 141–151 (2015). [CrossRef]
24. M. Mujat, B. Park, B. Cense, T. Chen, and J. de Boer, “Autocalibration of spectral-domain optical coherence tomography spectrometers for in vivo quantitative retinal nerve fiber layer birefringence determination,” J. Biomed. Opt. 12(4), 041205 (2007). [CrossRef]
25. J. de Boer and T. Milner, “Review of polarization sensitive optical coherence tomography and Stokes vector determination,” J. Biomed. Opt. 7(3), 359 (2002). [CrossRef]
26. P. Tang and R. K. Wang, “Polarization sensitive optical coherence tomography for imaging microvascular information within living tissue without polarization-induced artifacts,” Biomed. Opt. Express 11(11), 6379–6388 (2020). [CrossRef]
27. B. H. Park, M. C. Pierce, B. Cense, and J. F. de Boer, “Real-time multi-functional optical coherence tomography,” Opt. Express 11(7), 782–793 (2003). [CrossRef]
28. B. H. Park, C. Saxer, S. M. Srinivas, J. S. Nelson, and J. F. de Boer, “In vivo burn depth determination by high-speed fiber-based polarization sensitive optical coherence tomography,” J. Biomed. Opt. 6(4), 474–479 (2001). [CrossRef]
29. B. Cense, M. Mujat, T. C. Chen, B. Hyle Park, and J. F. de Boer, “Polarization-sensitive spectral-domain optical coherence tomography using a single line scan camera,” Opt. Express 15(5), 2421–2431 (2007). [CrossRef]
30. T. Y. P. Chui, D. A. VanNasdale, and S. A. Burns, “The use of forward scatter to improve retinal vascular imaging with an adaptive optics scanning laser ophthalmoscope,” Biomed. Opt. Express 3(10), 2537–2549 (2012). [CrossRef]
31. B. R. White, M. C. Pierce, N. Nassif, B. Cense, B. H. Park, G. J. Tearney, B. E. Bouma, T. C. Chen, and J. F. de Boer, “In vivo dynamic human retinal blood flow imaging using ultra-high-speed spectral domain optical Doppler tomography,” Opt. Express 11(25), 3490–3497 (2003). [CrossRef]
32. B. Hyle Park, M. C. Pierce, B. Cense, S.-H. Yun, M. Mujat, G. J. Tearney, B. E. Bouma, and d. J. F. Boer, “Real-time fiber-based multi-functional spectral-domain optical coherence tomography at 1.3 µm,” Opt. Express 13(11), 3931–3944 (2005). [CrossRef]
33. B. Cense, T. C. Chen, B. H. Park, M. C. Pierce, and J. F. de Boer, “Thickness and birefringence of healthy retinal nerve fiber layer tissue measured with polarization-sensitive optical coherence tomography,” Invest. Ophthalmol. Visual Sci. 45(8), 2606–2612 (2004). [CrossRef]
34. D. L. Streiner and G. R. Norman, ““Precision” and “accuracy”: two terms that are neither,” J. Chronic Dis. 59(4), 327–330 (2006). [CrossRef]
35. Y. Ouyang, Q. Shao, D. Scharf, A. M. Joussen, and F. M. Heussen, “An easy method to differentiate retinal arteries from veins by spectral domain optical coherence tomography: retrospective, observational case series,” BMC Ophthalmol. 14(1), 66 (2014). [CrossRef]
36. G. J. Ughi, M. G. Marosfoi, R. M. King, J. Caroff, L. M. Peterson, B. H. Duncan, E. T. Langan, A. Collins, A. Leporati, S. Rousselle, D. K. Lopes, M. J. Gounis, and A. S. Puri, “A neurovascular high-frequency optical coherence tomography system enables in situ cerebrovascular volumetric microscopy,” Nat. Commun. 11(1), 3851 (2020). [CrossRef]
37. B. A. Brown, H. Williams, S. J. George, and R. A. Khalil, “Chapter Six - Evidence for the Involvement of Matrix-Degrading Metalloproteinases (MMPs) in Atherosclerosis,” in Progress in Molecular Biology and Translational Science147 (Academic Press, 2017), pp. 197–237. [CrossRef]
38. T. Y. P. Chui, T. J. Gast, and S. A. Burns, “Imaging of vascular wall fine structure in the human retina using adaptive optics scanning laser ophthalmoscopy,” Invest. Ophthalmol. Visual Sci. 54(10), 7115–7124 (2013). [CrossRef]
39. Y. Muraoka, A. Tsujikawa, K. Kumagai, M. Akiba, K. Ogino, T. Murakami, Y. Akagi-Kurashige, K. Miyamoto, and N. Yoshimura, “Age- and hypertension-dependent changes in retinal vessel diameter and wall thickness: an optical coherence tomography study,” Am. J. Ophthalmol. 156(4), 706–714.e2 (2013). [CrossRef]
40. T. P. Zhu, Y. H. Tong, H. J. Zhan, and J. Ma, “Update on retinal vessel structure measurement with spectral-domain optical coherence tomography,” Microvasc. Res. 95, 7–14 (2014). [CrossRef]
41. G. Michelson, S. Wärntges, D. Baleanu, J. Welzenbach, A. Ohno-Jinno, P. Pogorelov, and J. Harazny, “Morphometric age-related evaluation of small retinal vessels by scanning laser Doppler flowmetry: Determination of a Vessel Wall Index,” Retina 27(4), 490–498 (2007). [CrossRef]
42. K. Otsuka, M. Villiger, S. K. Nadkarni, and B. E. Bouma, “Intravascular polarimetry: clinical translation and future applications of catheter-based polarization sensitive optical frequency domain imaging,” Frontiers in Cardiovascular Medicine 7(2020).
43. B. Cense, M. Reddikumar, and J. Cervantes, “Ultra-high resolution polarization-sensitive optical coherence tomography for imaging of the retinal nerve fiber layer,” SPIE Technologies and Applications of Structured Light (SPIE, 2017), Vol. 10251.
44. R. Maddipatla, J. Cervantes, Y. Otani, and B. Cense, “Retinal imaging with optical coherence tomography and low-loss adaptive optics using a 2.8-mm beam size,” J. Biophotonics 12(6), e201800192 (2019). [CrossRef]
