The Source of Moving Particles in Parafoveal Capillaries Detected by Adaptive Optics Scanning Laser Ophthalmoscopy

Abstract

Purpose.: To investigate the source of moving particles in parafoveal capillaries detected by adaptive optics scanning laser ophthalmoscopy (AO-SLO).

Methods.: AO-SLO videos were acquired from the parafoveal areas of eyes of healthy subjects. The gray-scale values inside and outside the moving particles were measured and compared. Thereafter, successive frames of the captured videos were analyzed under higher magnification to detect changes in the gray values of bright spots inside the capillaries, before and during passage of the particles. Simultaneously, changes in the gray values of areas without the bright spots were measured for comparison. Then, the authors analyzed the packing arrangements of the bright spots in the particles, and measured the particle velocity using spatiotemporal images of the target capillary.

Results.: There were no significant differences in the gray values between the moving particles and the cone mosaic outside the parafoveal capillaries adjacent to the particles. The gray value of the bright spots in the dark shadow of the vessels increased when the particles passed through, while the dark areas without bright spots remained dark. There were no significant differences in the packing arrangements between the bright spots and surrounding cone mosaic. Further, the concordance rate of packing arrangements of bright spots between two consecutive moving particles in the capillary was 95.8%. The mean particle velocity was 1.34 ± 0.42 mm/s.

Conclusions.: The particles moving in the capillaries are suggested to be reflections of photoreceptor aggregates that pass through circulating transparent objects such as leukocytes or plasma gaps.

Leukocytes play an important role in the pathogenesis of retinal vascular disorders such as diabetic retinopathy, retinal vein occlusion, and uveitis. Many studies in which histologic ^{1 –3} and serial acridine orange leukocyte fluorography ^{4 –9} were used have reported the involvement of leukocyte stasis (leukostasis) in capillary nonperfusion, endothelial cell damage, and increased vascular permeability after breakdown of the blood–retinal barrier in diabetic retinopathy and other diseases. Thus, leukocyte dynamics is considered to be important, and the ability to noninvasively investigate it in clinical practice has long been awaited. 

Thus far, several methods of evaluating the transit of leukocytes in the human retina have been reported. One such method uses the blue field entoptic phenomenon, introduced by Riva and Petrig. ¹⁰ The phenomenon is the perception of entoptic images observed against bright blue illumination and is caused by the movement of leukocytes in the macular capillaries. ¹¹ Previous studies have noninvasively evaluated the velocity of leukocytes and the hemodynamics of retinal capillaries by matching the subject's entoptic images with the leukocyte motion simulated on a computer screen. ^{12 –15} However, this method of evaluation is subjective. Paques et al., ¹⁶ on the other hand, investigated leukocyte dynamics with fluorescein-labeled autologous leukocytes. Although this is an objective evaluation technique, the monitoring time was not sufficient, and the method was invasive. Objective and less invasive methods for studying leukocyte dynamics in the human retina remain to be established. 

Recent studies have shown that noninvasive imaging of retinal cells such as photoreceptors and leukocytes in patients is possible with confocal adaptive optics scanning laser ophthalmoscopy (AO-SLO). ^{17 –23} AO-SLO is an SLO-equipped adaptive optics technology that provides high-resolution and high-contrast retinal images by correcting ocular aberrations. ¹⁸ Martin et al. ¹⁷ used AO-SLO noninvasively to demonstrate leukocyte movement visualized as bright particles in dark parafoveal capillaries; further, their pulsatility was measurable without the use of contrast dyes. ²³ AO-SLO provides a unique opportunity to study leukocyte dynamics in patients in an objective and noninvasive manner. However, the source of the high-intensity particles, determined to be leukocytes by AO-SLO imaging, moving in the dark vessels has not yet been identified. Considering the optical properties of blood plasma and blood cells, ²⁴ we hypothesized that parafoveal capillary leukocytes, which interrupt the column of erythrocytes, and plasma gaps between erythrocytes permit the AO-SLO imaging laser light to reach the photoreceptors beneath the vessels, facilitating detection of the light reflected from the photoreceptors at locations corresponding to the presence of leukocytes or plasma gaps, when the imaging depth is focused on the photoreceptor layer (Fig. 1). Here, we tested this hypothesis using our prototype AO-SLO system to identify the source of particles flowing in the parafoveal capillaries and to analyze leukocyte dynamics in normal retinal parafoveal capillaries. 

Figure 1.

 

Detection of the bright reflected light of photoreceptors in locations corresponding to the presence of leukocytes and plasma gaps. (A) Representation of the optical properties of blood plasma and blood cells in parafoveal capillaries. Parafoveal capillary leukocytes interrupting the column of erythrocytes that blocks or weakens the laser of the AO-SLO or plasma gaps between erythrocytes that permit the laser light to transmit to the photoreceptors beneath the vessels, enabling the detection of the bright reflected light of photoreceptors at locations corresponding to the presence of leukocytes or plasma gaps. (B) Leukocytes or plasma gaps (red frames) moving in the dark vessels are detected in the AO-SLO image as reflected light from photoreceptors (white disks).

Method

This study was approved by the Institutional Review Board and the Ethics Committee at Kyoto University Graduate School of Medicine and performed in accordance with the Declaration of Helsinki. Informed consent was obtained from each participant after a detailed explanation of the nature and possible consequences of the study procedures. 

Subjects

AO-SLO videos were acquired from the parafoveal areas of the left eyes of five healthy subjects without a history of prior ocular or systemic diseases (mean age ± SD, 37.0 ± 11.4 years; range, 27–53 years). 

AO-SLO Imaging

We developed a novel AO-SLO system with a high wavefront correction efficiency, using a dual liquid-crystal phase modulator (LCOS-SLM; X10468-02, Hamamatsu, Hamamatsu City, Japan), as described previously. ²⁵ Briefly, the imaging wavelength was 840 ± 25 nm, and the wavelength of beacon light for the measurement of wavefront aberrations was 760 ± 5 nm. The imaging light was set at 330 μW and the beacon light at 40 μW, by calculating the incident power of both light sources in accordance with the safety limits set by the American National Standards Institute. ²⁶ The videos were acquired at 32 or 64 frames per second. The scan area was either 1.2 ° × 1.2° or 1.4 ° × 2.8° at the retina and was sampled at 400 × 400 pixels or 200 × 400 pixels, respectively. 

Video Processing

Desinusoiding.

To generate a high-resolution AO-SLO image, it is necessary to calculate its pixel value at regular intervals. The resonant scanner changes its position sinusoidally and obtains reflective value at regular time intervals. Thus, the resonant scanner moves faster near the image center than near the edge. Accordingly, it collects few reflective values near the center and many near the edge of the image. We assumed that the reflective value acquired by the resonant scanner at regular time intervals is the integrated reflective value recorded while the resonant scanner moves across each pixel position. We calculated each pixel value by weighing the integration values based on the distance from the positions in which the integrated reflective values were acquired. 

Video Stabilization.

Raw videos were corrected for scanning distortions and stabilized to correct for eye motion using bUnwarpJ, an algorithm for elastic and consistent image registration developed as an ImageJ plugin (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). ²⁷ A fixed reference frame was used for image warping by bUnwarpJ. 

AO-SLO Movie Acquisition

AO-SLO videos were recorded for 2 to 4 seconds per scan area, and 10 to 25 scan areas were collected per subject, to cover the parafoveal areas. The average time taken to record a set of videos was approximately 15 minutes, excluding the time for dilation of the pupils. AO-SLO imaging was performed by focusing on the photoreceptor layer to enable detection of the cone mosaic pattern. 

Quantitative Image Analysis of Moving Particles

All the digital images, except for assessment of the spatial organization of the cone mosaics, were processed by a single operator (AU), who used the public-domain software ImageJ. The gray scale ranged from 0 (black) to 255 (white). 

Comparison of Gray-Scale Values between the Moving Particles and the Areas of Cone Mosaic Adjacent to the Particles.

The transparency of the moving particles was determined by comparing the gray values of the areas corresponding to the particles and to the cone mosaic adjacent to the particles, as follows (Fig. 2): Regions of interest (ROIs) were set for the areas inside the moving particles, and the same ROIs were applied to the cone mosaic areas outside the capillaries adjacent to each particle so that the same number of pixels was analyzed. Then, the average gray value within the ROI (i.e., the sum of the gray values of all the pixels in the ROI divided by the total number of pixels) was calculated for both the inside and outside areas of the particles. Five individual capillaries were measured per subject. For each capillary, the gray values of three different moving particles and the three areas adjacent to them were measured, yielding 150 measurements. 

Figure 2.

 

Comparison of the intensity between the moving particle and the area surrounding it. (B–D) Three consecutive frames corresponding to the area outlined in white in (A) showing a white particle (inside the circle) moving through a capillary. Scale bar, 100 μm. (D) Quantitative image analysis of the moving particle. The same regions of interest (ROIs) were applied to the particle (red-dotted outline) and to the area surrounding it (yellow-dotted outline). Then, average gray values within the ROIs of the particle and its surrounding area were compared. The intensity of the particle appeared equal to that of the surrounding area.

Changes in the Gray Values of Bright Spots and Dark Areas inside the Capillaries before and during Passage of the Moving Particles.

We analyzed successive frames under higher magnification to evaluate changes in the gray values of bright spots inside the capillaries before and during the passage of moving particles (Fig. 3). To be more specific, we set ROIs for single bright spots and measured the variations in the gray value caused by the moving particles. Simultaneously, we measured changes in the gray values of areas where reflections of the bright spots were absent, and the same ROIs were applied and measured for comparison. Five individual capillaries were measured per subject. For each capillary, the gray values of three different ROIs set on the bright spots and three different ROIs set on the areas without reflection of the bright spot were measured, yielding 150 measurements. 

Figure 3.

 

Changes in the intensity of bright spots inside the capillary before and during the passage of moving particles. (A) A frame from sequential frames of an AO-SLO video with a focus on the cone mosaic pattern. A bright particle moving in the dark shadow of the vessel is observed (white outline). Scale bar, 100 μm. (B, C) Magnified views of the area outlined in (A). The ROI was set for a single bright spot (red circle), and the changes in the gray value before (B) and during passage (C) of the particle were evaluated under higher magnification. Simultaneously, variations in the gray value of the areas where the bright spot was absent (red dotted circle) were measured for comparison.

Analysis of the Packing Arrangement of the Bright Spots Detected in the Moving Particles

To study the packing arrangement of the bright spots in the moving particles, we compared the spacing between the bright spots and the cones surrounding the capillaries. Li and Roorda ²⁸ reported an automated cone-labeling process carried out with an algorithm implemented in commercial software (MatLab; The MathWorks Inc., Natick, MA; with the MatLab Image Processing Toolbox; The MathWorks). We used this algorithm to automatically identify the cones and bright spots in the moving particles (see Fig. 6). ²² To estimate the density of cones or bright spots, we divided the number of cones or bright spots by the area of sampled region of each retina. During the measurement, accurate scan lengths were obtained by correcting the effect of magnification in each eye by using the adjusted axial length method devised by Bennett et al. ²⁹ To assess the spatial organization of the cones and bright spots, we examined the nearest-neighbor distances (NNDs) with a special software program that we wrote for the analysis. Furthermore, Voronoi domains associated with the cones or bright spots in each mosaic were examined. ^28,30,31 The domains were constructed for each spot by defining points in the regions that were closer to that spot than to any other spot in the mosaic. The NNDs were determined by calculating the minimum distances from the center of that spot to the centers of every other spot in the mosaic. Triangularly packed NNDs were calculated as expected for a perfectly triangularly packed mosaic with a density equal to that in each location. Five individual capillaries per subject were measured. Packing arrangements of two individual particles and two areas outside of the capillaries were measured for each capillary, yielding 100 measurements. 

Further, we analyzed differences in the packing arrangement between two consecutive moving particles to verify that the same spots are observed at every instance that a high-intensity particle passes through. After the same ROI was set on two consecutive particles in the capillary, two individual arrangements of automatically detected bright spots in the ROI were compared (see Fig. 6). Two experienced masked examiners (SO, KT) evaluated the different packing arrangements between two particles. In the case of disagreement, the results were discussed until consensus was reached. The concordance rate of packing arrangements between two particles was calculated as follows:  

 

where S_2A, S₁, and S₂ are the number of all spots detected in the ROI set on the second particle, the number of spots detected only in the ROI set on the first particle, and the number of spots detected only in the ROI set on the second particle, respectively. Five individual capillaries were evaluated per subject, yielding 25 pairs of particles. 

Velocity of the Moving Particles

We measured the particle velocity in the parafoveal capillary by using spatiotemporal images of the target capillary obtained by reslicing the sequential frames (Fig. 4), ^{32 –36} to confirm whether the particle behavior detected by our AO-SLO system was in accordance with the behavior of leukocytes reported in previous studies. ^{4,10,17,23,34,36} After the frames were resliced along the line set on the target vessel, the velocity of the moving particles was obtained by calculating the reciprocal of the slope of the white line depicted in the spatiotemporal image with the length of the line on the horizontal axis and the frame number on the vertical axis. The retinal field angle was converted to the actual distance on the retina on the basis of the axial length of each subject's eye. ²⁹ Five individual capillaries per subject, which were selected randomly, were measured. For each capillary, velocities of two individual successive particles were measured, yielding 50 measurements. 

Figure 4.

 

Velocity measurement of a moving particle. (A) Stacked sequential series of AO-SLO images. (B) A spatiotemporal image generated from (A) by reslicing the frames along the line set on the target vessel (A, red two-headed arrow). The velocity of the moving particle was obtained by calculating the reciprocal of the slope (Δx/Δt) of the white line (white arrowheads) depicted in the spatiotemporal image, graphed with the length of line on the horizontal axis and the frame number on the vertical axis. Thick and high-contrast lines (white arrowheads) as well as thin and dense white lines (red arrowheads) are depicted in the spatiotemporal image. Note that dark areas in the direction of the vertical axis are directly beneath the high-intensity slope ().

Statistical Analysis

All values are presented as the mean ± SD. Paired t tests were used to determine the changes, differences in the gray values, and comparisons of factors in relation to spatial arrangement of bright spots or cone mosaic (density, percentage of hexagonal Voronoi polygons, NNDs, triangularly packed NNDs, and ratio of the observed mean NND to the triangularly packed mosaic NND). P < 0.05 indicated statistical significance (all statistical analyses. StatView, ver. 5.0; SAS Inc., Cary, NC).. 

Results

Video Images of the Moving Particles within the Parafoveal Capillary

With the aid of our prototype AO-SLO system, we successfully obtained videos of the cone mosaic pattern on the retina and of particles moving in the parafoveal capillary (Supplementary Movies S1, S2). When AO-SLO imaging was performed by focusing on the cone mosaic, parafoveal capillaries were seen as dendritic dark shadows in the bright cone mosaic. The particles were observed to move rapidly within the parafoveal capillary shadow (Fig. 2). When the parafoveal capillaries were observed in magnified view, the moving particles appeared to consist of bright spots that were similar in size and brightness with individual photoreceptors of cone mosaic (Figs. 1B, 3). In addition, a spot with decreased brightness (presumably with slight reflection from a single photoreceptor) was observed at the same location as the bright spot, both before and after the passage of the particles. 

Quantitative Image Analysis of the High-Intensity Particles

There were no significant differences in the gray values between the areas corresponding to the moving particles or to the cone mosaic outside the parafoveal capillaries adjacent to the particles (Table 1). The gray values of the spots within the dark area corresponding to parafoveal capillaries seen in magnified view (Fig. 5) were significantly higher during the passage of particles than before the passage, whereas there were no significant changes in the gray values of the dark areas within capillaries lacking reflections of the cone photoreceptors (Table 2). 

Table 1.

 

Comparison of Gray Values between the Moving Particles and the Areas of Cone Mosaic Adjacent to the Particles

Subject	Particle	Cone Mosaic Area Adjacent to the Particle	P
A	131.07 ± 19.16	132.46 ± 15.81	0.809
B	125.61 ± 16.36	132.43 ± 24.56	0.149
C	84.20 ± 17.19	87.13 ± 17.00	0.627
D	125.84 ± 26.17	121.61 ± 19.03	0.399
E	115.50 ± 13.66	112.68 ± 8.93	0.260
Average	116.44 ± 25.10	117.26 ± 24.19	0.703

 Data are expressed as the gray-scale value ± SD.

Figure 5.

 

Moving particle in the parafoveal capillary detected as the photoreceptor aggregates. (A) Color fundus photograph. (B) Magnified view of outlined area in (A). (C) Montage of AO-SLO images corresponds to the outlined area in (B). (D–G) Four consecutive frames of the outlined area in (C) showing a white particle (circle) moving through a capillary. (H–J) Magnified views of (D–F). Note that the intensity of the spot (presumably an area with reflection from a single photoreceptor) changes from dark (dotted circle) to bright (solid circle), whereas that of the areas without photoreceptors remains low.

Table 2.

 

Changes in the Gray Values of Bright Spots and the Dark Areas Detected Inside the Capillaries Before and During Passage of the Moving Particles

Subject	Bright Spot			Dark Area
	Before Passage	During Passage	P	Before Passage	During Passage	P
A	104.62 ± 28.30	162.59 ± 34.59	<0.0001	43.15 ± 7.44	46.11 ± 8.27	0.286
B	80.49 ± 27.54	132.84 ± 51.13	<0.0001	44.36 ± 20.15	41.38 ± 19.82	0.303
C	69.81 ± 13.51	141.65 ± 27.96	<0.0001	26.78 ± 11.65	28.20 ± 11.20	0.449
D	92.06 ± 32.87	143.39 ± 42.52	<0.0001	43.59 ± 13.24	44.07 ± 9.40	0.880
E	103.77 ± 17.20	143.89 ± 31.04	<0.0001	50.611 ± 8.12	51.91 ± 12.96	0.576
Average	90.15 ± 27.81	144.87 ± 38.60	<0.0001	41.70 ± 14.91	42.33 ± 14.91	0.580

 Data are expressed as the gray-scale value ± SD.

Figure 6 and Table 3 show the results of the Voronoi and nearest-neighbor analyses. There were no significant differences between the arrangements of bright spots detected in the moving particles and surrounding cones with regard to density, percentage of hexagonal Voronoi polygons, NNDs, and triangularly packed NNDs, suggesting that the packing arrangement of the bright spots is appropriate, given the retinal eccentricity. Further, the concordance rate of the packing arrangements of bright spots between two consecutive particles was 95.8%. Of the 25 pairs of moving particles, 12 were discordant, and the mean number of discordant spots was 0.8 (range, 0–3). The discordance was considered to be due to the failure of the algorithm to automatically identify the bright spots. This high concordance rate suggests that the same spots are seen every time a particle passes through. Taken together, these results encourage us to speculate that these bright spots are not structures within the moving object (e.g., internal leukocyte structures) but are cone mosaics. 

Figure 6.

 

Spatial arrangement of the bright spots detected inside particles flowing in the dark shadow of a capillary[b]. (A–D) Four consecutive frames show a particle (inside the circle) moving through a capillary. Scale bar, 50 μm. (E) Processed image of (B). Cones and bright spots in the moving particle (white dotted outline) are automatically labeled (red dots). There are no apparent differences in the density of cones and bright spots. (F) Color–coded Voronoi diagram of (E). The colors indicate the number of sides of each Voronoi polygon (pink, 4; blue, 5; green, 6; yellow, 7; and orange, 8). Large regions of the six-sided green polygons indicate a regular triangular lattice, whereas the other colored regions indicate points of disruption in the hexagonal packing of the cone mosaic. The number of six-sided Voronoi polygons in the moving particle (outlined in white) is similar to that in the area of cone mosaic, whereas it is smaller in the areas corresponding to the dark shadow of the capillary. (G, H) Magnified view of (E) and (F). (I) Processed image of a frame acquired 1.67 seconds after frame (B). Moving particle (outlined by white dotted line) is detected in almost the same location as the particle detected in (E). (J) Color-coded Voronoi diagram of (I). (K) Magnified view of (I). The arrangement of bright spots (1–12) in the ROI (outlined in yellow) set in the moving particle white dotted outline) corresponds to that of the bright spots detected in same ROI in (G). (L) Magnified view of (J). The Voronoi domain is similar to that in (H), but there is no full agreement on this detail because of some minor differences in the automatic detection of the bright spots.

Table 3.

 

Spatial Arrangement of the Bright Spots Detected Inside Particles Flowing in the Dark Shadow of the Capillaries

Variable	Bright Spots Inside Moving Particles	Cone Mosaic Outside the Shadow of the Capillary	P
Density, spots/mm2 or cones/mm2	21,977 ± 5,065	22,155 ± 4,512	0.718
Percent of hexagonal Voronoi polygons	45.8 ± 14.1	48.3 ± 8.5	0.183
Observed NND, μm	5.45 ± 0.66	5.42 ± 0.65	0.469
Triangularly packed NND, μm*	7.40 ± 0.93	7.34 ± 0.69	0.386
Ratio†	0.74 ± 0.05	0.74 ± 0.04	0.878

 Data are expressed as the mean ± SD.

*  Calculated NND as expected for a perfectly triangularly packed mosaic with a density equal to that observed for each location.

†  Ratio of the observed mean NND to the triangularly packed mosaic NND. Lower ratios indicate a larger departure from a perfectly packed mosaic.

These results suggest that the particles moving in the capillaries are not substantial, but consist of highly reflected light from photoreceptor aggregates through transparent moving objects such as leukocytes or plasma gaps (Figs. 1, 5). 

Velocity Measurement of the Moving Particles

The velocities of the high-intensity particles ranged from 0.73 to 2.36 mm/s for all subjects, and the mean velocity was 1.34 ± 0.42 mm/s. Intriguingly, in the spatiotemporal images, we found dark areas in the target capillaries along a certain length in the vertical axis following the high-intensity slope and corresponding to the trajectory of the moving particles (Fig. 4). In some raw videos, we detected regions that were darker than the vessel shadow, after the passage of the particles (Supplementary Movies S1, S2). 

Discussion

The internal structure of the bright moving particles in the dark vessels detected by AO-SLO had a feature similar to cone photoreceptors, suggesting that the moving particles consist of highly reflected light from photoreceptor aggregates through transparent moving objects. Leukocytes are candidates that permit the AO-SLO imaging laser light to reach the photoreceptors. In contrast to erythrocytes, leukocytes are water-clear blood cells with fairly low absorptivity, thus permitting the AO-SLO imaging laser light to reach the photoreceptors. ²⁴ Therefore, the AO-SLO laser is thought to transmit the leukocyte body with the least decay in laser intensity, resulting in detection of leukocytes as background illumination of photoreceptors. A magnified view of the bright spots in the shadow of the capillary vessels revealed that the cone photoreceptors lit up in a sequence, resembling a microscopic electronic billboard. 

In our videos (Supplementary Movies S1, S2), the moving particles were readily detected in the capillaries; however, they were scarcely detected in larger vessels such as the terminal artery or collecting venules. We believe that an explanation could be that erythrocytes can easily surround the leukocyte in vessels of large caliber, rendering it difficult for the AO-SLO imaging laser to pass through the vessels. In contrast, leukocytes occupy the entire space of the lumen in capillaries, and blood cells including erythrocytes and leukocytes move in a single file, resulting in the easier detection of leukocytes as an aggregate of reflection from photoreceptors in AO-SLO imaging.^17,37  

We recorded videos with a focus on the cone mosaic as previously reported.^17,23,38 This focusing appeared most effective for imaging moving leukocytes, because the reflected light from the photoreceptors corresponds to the presence and locations of leukocytes in AO-SLO imaging. Videos with a focus on the capillary layer (Supplementary Movie S3) revealed the presence of numerous, moving, high-intensity particles in the capillaries that were smaller than the particles detected in the videos with a focus on the photoreceptor layer (Supplementary Movies S1, S2). These high-intensity particles were numerous and very close to each other, and their velocity and directions of flow could not be detected. Since our videos were acquired by a confocal optics system, we speculate that these numerous particles, which showed large scattering coefficients compared with blood plasma and leukocytes, were erythrocytes.²⁴ Although, at this focus, blood plasma and leukocytes appear as dark particles due to their low reflectivity, we could not detect such particles in the videos. We think that this contradictory appearance is due to the imperfection of the confocal optic system. Briefly, imaging laser light reflected from the retinal layers near the capillary layer was also detected as defocused structures, even when the focus was on the capillary layer. A bright, dotted background derived from photoreceptors both outside and inside the capillaries was visualized when the system was focused on the capillary layer (Supplementary Movie S3), rendering it difficult to distinguish between bright dots from photoreceptors through blood plasma and leukocytes and those from erythrocytes. In contrast, when focused on the photoreceptor layer, erythrocytes were visualized as shadows cast on the layer beneath the capillary layer, and blood plasma and leukocytes appeared as bright moving particles. Thus, it seems reasonable to focus on the photoreceptor layer for the purpose of distinguishing between the blood components. 

Blood plasma is another candidate through which the light reflected from photoreceptors (840 nm) can pass with minimal decay, resulting in the detection of plasma as bright moving particles, although this signal may be slightly weaker than that of leukocytes. Since it is important to distinguish between leukocytes and the plasma gap in AO-SLO imaging, in the study of leukocyte dynamics, it is necessary to establish a method of identifying these moving particles. Previously, Tam et al. ³⁴ described several unique features in spatiotemporal images generated from the videos processed with motion contrast enhancement, aiding in the characterization of the moving particles. In that study, they defined the spatiotemporal plots derived from leukocytes as thick, high-contrast, sparse, and unidirectional trajectories, and the spatiotemporal plots derived from plasma gaps as thin and dense trajectories. Although we used spatiotemporal images generated from videos without motion contrast enhancement, our images appeared to detect spatiotemporal plot patterns that were similar with their findings (Fig. 4). Since the trajectories that met the criteria of plasma gaps were too faint to use for measurements of the particle velocities, we could use only those trajectories that met the criteria of the leukocytes in our study. 

Our spatiotemporal image analysis revealed a dark band directly beneath the high-intensity slope. Careful observation of the raw videos revealed a region darker than the vessel shadow that occurred closely after the particles passed through. This region corresponded to the dark band of the spatiotemporal image (Supplementary Movie S2). We speculated that the dark bands in the spatiotemporal images were due to the biased distribution of erythrocytes upstream and downstream of the leukocytes, as reported previously. ^4,37 Because of their smaller deformability and larger volume than erythrocytes, ³⁹ leukocytes in capillaries are slower than erythrocytes. Therefore, erythrocytes are packed right upstream of the leukocyte, whereas erythrocyte-depleted regions are formed downstream. The dark region in both spatiotemporal images and videos may correspond to the aggregate of erythrocytes upstream of the leukocytes, which block the AO-SLO laser; we speculate that this observation corresponds to the dark tail observed in the blue field entoptic phenomenon. ¹⁰ Thus, the tadpole tail–like structure appearing in our videos may help in distinguishing leukocytes from the plasma gap. On the other hand, it follows from this biased distribution of erythrocytes that the high-intensity signals may be not only a single file of leukocytes or plasma gaps but also a plasma gap–leukocyte complex, since erythrocyte-depleted regions that may make the bright, moving particles longer than the original length of the leukocytes are formed immediately downstream of the leukocytes. In future studies, we will evaluate the clinical importance of the presence of these findings in retinal circulation. 

The particle velocities were calculated using spatiotemporal images generated from raw videos in which high-intensity and thick lines represented the particle trajectories. The mean particle velocity in retinal capillaries that we found in our study was in good agreement with the data of Yang et al., ⁴⁰ who reported a mean leukocyte velocity of 1.37 mm/s determined with SLO and fluorescein angiography (FA), and with the data of Paques et al., ¹⁶ who reported a mean velocity of 1.43 mm/s determined with FA using fluorescein-labeled autologous leukocytes. Grunwald et al. ¹⁴ and Arend et al. ⁴¹ used the blue field entoptic phenomenon and reported values of 0.89 and 0.75 mm/s, respectively, which are lower than the values in our study. As a reference, Nishiwaki et al. ⁴ used acridine orange staining and reported the mean leukocyte velocity in monkeys to be 0.92 mm/s. Moreover, Martin et al. ^17,23 measured the mean leukocyte velocity directly from AO-SLO videos and reported it to be 1.37 mm/s, which is in excellent agreement with our results. However, our values were considerably lower than the corresponding data of Tam et al., ³⁴ who reported a value of 1.80 mm/s by using spatiotemporal images generated from AO-SLO videos acquired from one subject. The reason for this difference is unclear; however, because the particle velocity in our study was variable (0.73–2.36 mm/s), measurements can vary across different subjects and locations of measurements. Therefore, results from one subject would be likely to suffer more from this variability. Another possible reason for the difference is the lack of evaluation of pulsatility in our study. Measurement of pulsatility would reveal the cyclic changes in flow velocity, which would help in calculating the appropriate mean velocity. ^23,42,43 In the present study, we could not assess the pulsatility because our system requires an internal memory to allow storage of the obtained video data and restricts the size of the video to a few seconds, and this could be a limitation of our study. The lack of correction of error due to raster scanning also may be a reason for the difference in velocities. As reported by Tam and Roorda, ³⁵ we have to consider the relationship between the direction of moving objects and the direction of the raster scan to improve the method's accuracy in determining object velocity. Our results were obtained using spatiotemporal images without slope modification to reduce the error associated with raster scanning. Although evaluations of pulsatility and raster scan error are indispensable for studying blood flow velocity in the retinal circulation, we believe that excluding this parameter did not affect our main findings that the high-intensity particles were reflections of photoreceptor aggregates. 

Voronoi and nearest-neighbor analyses other than cone density were used for the comparison of spacing between the bright spots in the moving particles and cone mosaic around the capillaries. The percentage of hexagonal Voronoi polygons that we found compares well to the data reported by Li and Roorda, ²⁸ who found 40% to 50% in the retinal region at 0.17° to 5.00° from center of the fovea, which covers the parafoveal region analyzed in our study. Ooto et al. ²² reported the percentage of hexagonal Voronoi polygons to be 38.4% at 0.5 mm and 53.8% at 1.0 mm from center of the fovea, which was in good agreement with our data. They also reported a ratio of the observed mean NND to the triangularly packed mosaic NND that was similar to that in our data, where a lower ratio means a larger departure from a perfectly packed mosaic. They reported a ratio of 0.79 at both 0.5 and 1.0 mm from the center of the fovea. Finally, our data of cone density compares well with the data reported by these researchers. Li and Roorda ²⁸ reported 12,300 to 72,200 cones/mm² in the retinal region at 0.17° to 5.00° from center of the fovea. Ooto et al. ²² reported 31,960 cones/mm² at 0.5 mm and 15,116 cones/mm² at 1.0 mm from the center of the fovea. Taken together, these data all emphasize that the spacing that we found between spots or cones is appropriate, given the retinal eccentricity. 

In conclusion, the internal structure of the moving particles was similar to that in the cone mosaic, suggesting that these particles consist of photoreceptor aggregates and are transparent objects, such as leukocytes or plasma gaps, given their optical properties. Additional research on the characterization of blood components may enhance the application of AO-SLO to objectively and noninvasively study leukocyte behavior. 

Supplementary Materials

Text s1, DOCX - Text s1, DOCX 

Movie sm1, MOV - Movie sm1, MOV 

Movie sm2, MOV - Movie sm2, MOV 

Text sm3, MOV - Movie sm3, MOV 

Footnotes

 Presented in part at the annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida, May 2011.

Footnotes

 Supported in part by the Innovative Techno-Hub for Integrated Medical Bio-imaging Project of the Special Coordination Funds for Promoting Science and Technology from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan.

Footnotes

 Disclosure: A. Uji, None; M. Hangai, Topcon (F, C, R), Nidek (F, C, R), Heidelberg Engineering (R); S. Ooto, None; K. Takayama, None; N. Arakawa, None; H. Imamur, Canon (E); K. Nozat, Canon (E); N. Yoshimura, Topcon (F, C, R), Nidek (F, C, R), Canon (F, C, R), Heidelberg Engineering (R)