Marginal Corneal Vascular Arcades

Abstract

Purpose.: To determine the metrics of the marginal corneal vascular arcades (MCA).

Methods.: The MCA and filling pattern was investigated using indocyanine green dye angiography (ICGA) in the fellow eye of patients with treated unilateral keratitis. Images were acquired using a scanning laser ophthalmoscope. Five contiguous squares (100 pixels) were aligned beyond the inner row of vessels extending approximately 700 μm into the limbal region and spanning an arc length of approximately 4 mm of the peripheral cornea. Geometrical properties of the MCA were determined using programs written in a numerical computing environment.

Results.: A total of 17 patients (24–88 years) were included. Filling of the inferior corneal quadrant occurred first, followed by superior, nasal, and temporal quadrants. Mean area of a vascular loop of the MCA was 11.87 × 10⁻³ mm² (SD: 10.44 × 10⁻³ mm²) skewed (2.20) toward smaller sizes. Mean circumference of a vascular loop was 422.5 μm (SD: 218.7 μm) with major and minor axes of 158.9 μm and 90.8 μm. There were five (SD: 1.8) branches per loop with a segment length of 89.5 um (SD 163.8 μm). Vessels were tortuous (mean 0.19, SD: 0.16) with a fractal number of 1.51 (0.12). There were significant differences between subjects in vessel loop area (P = 0.003) and number of branches (P = 0.002). Speed of flow was circumferential along the innermost row and measured at 0.22 mm/s in one subject.

Conclusions.: The MCA comprise a network of branched interlinked elliptical loops supporting circumferential blood flow in the corneal periphery. There was no definable change in vascular pattern extending into the limbal region.

Introduction

The pathogenesis and features of diseases of the cornea may be dependent on the characteristics of its peripheral blood supply. The biomicroscopic features and vasculature of this region were described in detail by Graves, ¹ while Goldberg and Bron, ² Meyer and Watson, ³ and others ^4–7 described the fluorescein angiographic (FA) features. The palisades of Vogt that surround the cornea consist of a zone of thickened epithelial mucosa, most developed superiorly and inferiorly, whose epithelial rete pegs project down between a series of radially orientated fibrovascular ridges. ^2,8 The vessels of the palisades are organized in parallel rows, mainly superiorly and inferiorly where the palisades are most developed. ² The region of the limbus internal to the palisades and extending into the peripheral cornea contains the marginal corneal vascular arcades (MCA). ^1,2,8 There is little information about the anatomy and physiology of the MCA, particularly in the living human eye, possibly limited by the image acquisition and analysis systems. As has been demonstrated with corneal neovascularization (CoNV), many of the vessels—particularly the fine capillary networks that are seen with indocyanine green angiography (ICGA)—are not evident on color photography. ^9,10 The MCA comprise a system of vessels lying at a subepithelial level and represent the most centrally located vessels on the surface of the globe. ² Although the MCA officially belong to the cornea and overlie clear corneal tissue, they lie outside the peripheral edge of Bowman's layer and it has been suggested that their central tips provide a useful surface landmark for this feature. ⁸ The MCA arise from anterior branches of the episcleral circle as the latter pass forward, looping superficially to form the anterior conjunctival arteries. ² These episcleral arteries also supply the vessels of the palisades. ^2,3  

The MCA may be involved in disease processes that affect the peripheral cornea and may account for the characteristic features of these conditions. In addition, although CoNV may arise from the conjunctival or episcleral sources or limbal palisadal vessel, CoNV may also originate from the MCA. Therefore, characterization of the MCA is important for understanding the vascular supply of the cornea, and diseases and conditions that may involve the peripheral corneal vascular network such as marginal keratitis, peripheral corneal ulceration, or lipid deposition. 

The introduction ICGA—which is not subject to leakage and which provides excellent vessel delineation even in the presence of scars—together with increased magnification, computerized digital angiography, and developed image analysis, ^9,10 offers the opportunity to determine the metrics of the MCA in vivo and in greater detail than has previously been possible. 

Methods

Study Population

Patients who were being investigated for unilateral CoNV associated with keratitis using FA and ICGA were included. Images were taken of the eye with CoNV (FA) and of the fellow eye (ICGA) to characterize the MCA. 

Patients were excluded if they had contraindications to undergoing FA and or ICGA (patients with a known allergy to iodides and shellfish); were unable to fixate on a target or who have continuous eye movements such as nystagmus; were using topical eye medications to the healthy eye; had known diabetes, vasculitis, autoimmune, vascular malformations, or hematologic disease. Informed consent and ethics approval was obtained and the study was conducted in accordance with the tenets of the Declaration of Helsinki. 

Imaging of the MCA

Following a 5-mL intravenous injection of 5 mg/mL indocyanine green dye (Pulsion Medical Systems, Germany) into a peripheral arm vein, images were acquired using a scanning laser ophthalmoscope (Heidelberg Retina Angiograph; Heidelberg Engineering, Heidelberg, Germany) with a 32-diopter [D] focus incorporating automatic real-time (ART) software as previously described. ⁹ Five to 10 seconds after injection, videography was undertaken for 25 to 30 seconds using either a 30° field of view to encompass the whole 360° of the limbus, or 15° in Hi-Res mode to image a region of the MCA, followed by single frame ICGA photographs every 3 to 5 seconds for 3 minutes. Late ICGA images were taken at 5 and 10 minutes. 

To determine the metrics of the MCA, the best image of an area of one of the peripheral corneal quadrants of the fellow eye was selected using the same procedure as previously described for CoNV. ⁹ Briefly, images graded 3 or more by two independent observers (AT, SBK) based on qualitative grades from 0 to 4 (0 = no vessel discernible; 1 = poor vessel delineation; 2 = good vessel delineation; 3 = very good vessel delineation; 4 = excellent vessel delineation) were selected for analysis. ⁹ The resolution required to image the MCA and the curvature of the region limited the area that could be imaged and kept in focus to approximately 1 mm (meridionally) with an arc length of 4 mm. To determine the filling pattern of the peripheral cornea, images of the fellow eye of a patient were taken using lower magnification to capture the entire (360°) corneal periphery. 

Filling Pattern of the Peripheral Cornea

The onset of inner limbal and peripheral corneal fluorescence in each of four designated quadrants (nasal, superior, temporal, and inferior), together with the time period to filling of each respective quadrant, was measured independently by two observers (SBK, AT). 

Analysis of Images

The best available ICGA images of the marginal corneal arcades were selected based on criteria previously described for CoNV and exported in TIFF format for the purpose of quantitative analysis. ⁹ Five contiguous squares (100 pixels, 850 μm), spanning a circumferential arc length of approximately 4 mm, were aligned along the circumference of the inner border of the MCA so that the inner edge of each box extended beyond the innermost row of the MCA approximately 700 μm into the limbal region (Fig. 1A). The area spanned by each box—that is, 850 μm × 850 μm—defined the region of interest (ROI). The area of the vascular complexes in each square and geometrical properties of the vessels were determined on the selected images using programs written in a numerical computing environment (MATLAB R14; MathWorks, Inc., Natick, MA). In order to facilitate quantitative analysis and avoid bias caused by the borders of squares, a larger square window was used to crop the image for manual annotation. In the actual analysis of vessels and loops, only those that were within the ROIs were taken into account (any loops or vessels crossing the ROI border were excluded). This objective analysis was carried out on a computer (configurations: Windows XP Service Pack 2 [Microsoft Corp., Redmond, WA]; Intel Core 2 [Intel Corporation, Mountain View, CA], 2.66 GHz, and 3.25 GB of RAM). Due to the relatively low contrast and resolution of the image, it appeared difficult to automatically detect all the vessels in the image for the purpose of accurate analysis of their geometry. Instead, a semiautomated segmentation program (Live Vessel; available in the public domain at http://livevessel.cs.sfu.ca) was used to segment all the vessels visible to the observers. ¹¹ Live Vessel is an interactive vessel segmentation software based on multiscale vesselness filters and Livewire framework for locating vessel-like structures in 2-dimensional images. The image is opened in the provided graphic user interface (GUI) and the user moves a cursor from the start or seed point along the vessel with the aid of the software that automatically calculates the best vessel path from the seed point to the cursor position. As the user moves the cursor, the vessel is updated and displayed in the GUI. By repeating this process, the user can trace all the vessels of interest. The software can export the resulting vascular tree as a binary image (1 represents vessel pixels, 0 otherwise) for further processing. After appropriate training in the use of the program, two independent untrained and unbiased observers (AK, AB) traced all the vessels visible to them in each image (Figs. 1B–D). Each observer was limited to 3 hours per day with each test taking approximately 10 days to complete. In order to investigate intra- and interobserver agreement, both observers repeated the analysis with a minimum interval of 1 week between the first test and the second test. The box-counting approach was used to estimate the fractal complexity in each image. ¹² After manual segmentation of the vessels, a three-step automatic analysis process was used to quantify the geometrical property of the vasculature structure by adapting a well-established method. ^9,10 First, the centerlines of the segmented vessels were determined by a mathematical morphological thinning operation. The significant points (branch points and terminal points) were then identified and used to segment the vasculature into individual segments. 

Figure 1

 

MCAs. (A) Left eye. Example of locations of five subregions used for analysis. Size of box: 850 μm (arrow). (B) Vessel annotations of fourth subregion by two independent observers from two sessions, paired by observer.

For each vessel segment, the geometrical features, length, and tortuosity were measured and used to describe the overall properties for each patient. Tortuosity was calculated using the definition of the ratio between the length of the curved vessel and the square of the chord length of the vessel. ¹³ Vessel segment length was defined as the length of the vessel between two branch points or between a branch point and a terminal point. A terminal point was defined as where the vessel appeared to end blindly. An analysis of the defined region was undertaken in order to establish the number of loops enclosed in the region, area of each loop, minimum and maximum axis length of each loop, and the entire area of all vessel loops. The size of a vascular loop—that is, the area enclosed by each vascular loop (mm²)—was defined according to the number of pixels and pixel size in the x and y axes. From this analysis, the geometrical properties of the loops in the first row (innermost vascular loops) with respect to the cornea were also computed. Lengths were measured in μm and areas in 10⁻³ mm². The pixel size (8.5 μm/pixel) was estimated using a similar technique as described elsewhere. ⁹  

Analysis of Blood Flow in the MCA

In one patient, it was also possible to determine the direction and speed of flow in the innermost rows of the MCA from the video recordings. Direction of flow was observed from the movement of segments of fluorescence within the vessel. For the speed of flow, a hyperfluorescent spot was identified in one subject and the time taken for the spot to travel a specified distance was measured. A conversion from pixels to μm was calculated from reference points and from the horizontal corneal diameter (Fig. 2). Specifically, from the horizontal white-to-white corneal diameter, the resolution of the image was estimated and used to measure the distance between two landmarks as evidenced in both the infrared image and the ICGA image (Fig. 2). This distance was in turn used to estimate the pixel size of the ICGA image and the distance traveled of a hyperfluorescent spot over a certain time period. 

Figure 2

 

Inferior MCA. (A) Horizontal corneal diameter: 12 mm, 1060 pixels with a resolution 11.32 μm/pixel. (B) Distance between two locations in the inferior zone (94 pixels, 2.20 mm). (C) Distance between two landmarks (430.3 pixels), with a resolution of 5.1 μm/pixel.

Intra- and Interobserver Agreement

Observer agreement was measured using the intraclass correlation (ICC). All the statistical analyses were performed as appropriate using statistical software (SPSS version 17.0; SPSS, Inc., Chicago, IL). A nonparametric test (Kruskal-Wallis) was used to test for differences between subjects and P < 0.05 was considered statistically significant. 

Results

Patients

A total of 17 patients (11 males and 6 females, aged between 24 and 88 years) who were undergoing angiography for the investigation and management of unilateral (contralateral) CoNV associated with unilateral keratitis were included. A total of 15 patients with previous microbial keratitis (10 patients with bacterial ulcerative and 5 patients with recurrent herpetic keratitis) developed CoNV associated with scarring. They were in the remission stage (4–10 weeks following healing of the ulcer) and were receiving low-dose topical steroids to the affected eye. One patient had a chemical injury 3 months prior to imaging and one patient had received previous treatment with topical interferon for carcinoma in situ 6 months prior to imaging. None of the patients were using medication to the uninvolved eye. In 10 patients, the filling of the entire peripheral (360°) cornea was evident on the videographic images and these were analyzed to determine the pattern of filling of the peripheral cornea. High-resolution images of portions of the peripheral cornea were selected in seven patients to determine the characteristics of the MCA. 

Intraobserver and Interobserver Agreement

There was very good agreement both within and between observers for all the measured parameters. The mean (SD) interobserver and intraobserver ICC was 0.83 (0.08) for branch points, 0.77 (0.08) for fractal dimension, 0.75 (0.09) for vessel segment length, and 0.67 (0.11) for size of vessel loops. 

Characteristics of the MCA: Vessel Dimensions

The area of the peripheral cornea analyzed in each subject was approximately 850 μm by 4.00 mm. The geometric properties of the MCA for the seven patients are shown in Figure 3 and Tables 1 and 2. The mean (SD, median) number of vessel rows from the innermost row of the MCA that were included was five (SD 1.98, median 4) containing 59.4 (17.8, median 62) branch points. The mean vessel segment length was 89.5 μm (SD 163.8, median 75.0 μm) with a minimum of 14.5 μm and maximum of 717.0 μm. Vessels in the MCA were tortuous 0.19 (0.16, median 0.13) with five (1.8, median 5) branches per vascular loop and a mean fractal number of 1.51 (0.12, 1.49, minimum 1.32, maximum 1.69). The mean size of a vascular loop across the first five rows of the MCA from the cornea was 11.87 × 10⁻³ mm² with high variation (SD 10.44 × 10⁻³ mm²) and a skew (2.20) toward smaller sizes (minimum 0.07 × 10⁻³ mm², maximum of 70.08 × 10⁻³ mm²). The corresponding mean major and minor axes of a vascular loop were 158.9 μm and 90.8 μm, with a ratio close to 2 (1.8). For all the loops within the defined region, the mean length of the circumference was 422.5 μm (218.7 μm, median 377.3 μm). The mean size of the vessel loops in the first row—that is, the row closest to the cornea—was slightly bigger 14.45 × 10⁻³ mm² (Table 1), with a mean circumference of 464.4 μm (SD 244.4 μm, median 404.4 μm). 

Figure 3

 

Geometric properties of the MCAs. Histogram of variables measured from 35 subimages by observer 1 in session one. (A) Length of all vessel segments. (B) Tortuosity. (C) Area enclosed by each vascular loop. (D) Length of circumference of vascular loops.

Table 1

 

Geometric Properties of the MCAs

	Measurement
Marginal corneal arcades: all rows
Circumference of a vascular loop, μm	422.5 (218.7)
Segment length between branches, μm	89.5 (63.8)
Number (segments) branches per loop	5
Mean fractal number of a vascular loop	1.40
Median area of a vascular loop, × 10−3 mm2	8.71
Mean area of a vascular loop, × 10−3 mm2	11.87 (10.44)
Minor axis of a vascular loop, μm	90.8 (38.9)
Major axis of a vascular loop, μm	158.9 (81.4)
Calculated segment length of a vascular loop, μm	81.4
Ratio of calculated to measured length	0.91
Calculated length of circumference of loop, μm	399.6
Ratio of calculated to measured circumference	0.95
Calculated area of a vascular loop, × 10−3 mm2	11.33
Ratio of calculated to measured area of a vascular  loop	0.95
First row of the marginal corneal arcades
Mean area of a vascular loop, × 10−3 mm2	14.45 (12.51)
Median area of a vascular loop, × 10−3 mm2	10.19
Circumference of a vascular loop, μm	464.4 (244.4)
Calculated circumference of a vascular loop, μm	437.1
Minor axis of a vascular loop, μm	100.0 (42.0)
Major axis of a vascular loop, μm	173.3 (92.4)
Calculated area of a vascular loop, × 10−3 mm2	13.61
Ratio of calculated to measured area of a vascular  loop	0.93

Table 2

 

Geometric Properties of the MCAs—Patient Variation

Patient	Age, Sex	Zone	Rows, n	Branch Points of Vascular Loops, n	Area of a Vascular Loop Area, × 10−3 mm2
1	44, M	Temp	5	51.4 (12.3)	15.42 (4.17)
2	63, F	Sup	5	72.0 (16.9)	10.81 (2.41)
3	80, M	Sup	5	77.0 (13.9)	9.18 (1.73)
4	24, M	Inf	3	56.6 (16.1)	12.83 (1.94)
5	88, M	Inf	5	69.4 (5.9)	10.86 (1.33)
6	51, F	Inf	3	33.4 (6.1)	18.48 (4.30)
7	48, M	Inf	6	55.8 (8.4)	13.45 (2.75)

 The mean number of rows measured in each zone, mean (SD) number of branch points, and area of each loop for that zone in uninvolved eye. Sup, superior; inf, inferior; temp, temporal.

Patient Variation

The metrics of the MCA were measured in the inferior zone in four patients, superior zone in two patients, and temporal zone in one patient. There were significant differences (Kruskal-Wallis test) in the vessel loop area (P = 0.003), number of branches (P = 0.002) and number of rows (P = 0.029) between the seven patients (Table 3). For four of the subjects who had the inferior zone of their cornea imaged, the mean number of rows was 4, branch points 53.8 (9.18) with a mean area enclosed by a vascular loop of 13.91 (2.58) × 10⁻³ mm². 

Table 3

 

Filling Pattern of the MCAs

Corneal Quadrant	Time to Appearance of Fluorescence, s (SD)	Time Taken to Complete Filling of the Quadrant, s (SD)
Inferior	N/A	4.42 (3.01)
Superior	0.76 (0.57)	5.58 (2.88)
Nasal	2.39 (1.94)	6.84 (4.08)
Temporal	6.02 (2.69)	8.50 (5.13)

 Fellow eyes of 10 patients, mean age 44 years (SD: 22) with contralateral unilateral keratitis. Mean (SD) time taken in seconds for the appearance of fluorescence and completion of circumferential filling in each designated peripheral quadrant following the appearance of fluorescence in the inferior quadrant.

Approximation to an Ellipse

An ellipse gave a good approximation to the actual size of the measured vessel loop. Using πab for the area and π(a + b)|1+{[3((a − b)/(a + b))]²/[10 + Display Formula ]}| for the circumference ¹⁴ —where a and b are half the major and minor axes, respectively—gave a ratio of close to 1 for the actual to calculated area and circumference. 

Filling Pattern of the MCA

Due to the variability in the time taken following injection of the dye to the commencement of the video, it was not possible to determine the precise time following injection to the first appearance of fluorescence in the region of the MCA. In each of the 10 patients, there was a consistent pattern of filling with reference to the central zone of each of the four quadrants, commencing in the inferior quadrant followed by the superior almost at the same time, then the nasal and temporal zones. The mean (SD) time of appearance of the fluorescence in each of the midpoints of the quadrants for the 10 patients, taking the inferior quadrant as time zero, are shown in Table 3, together with the time taken for complete filling of the respective quadrant. 

Analysis of Blood Flow in the MCA

The circulation across the innermost row of the MCA appeared to be in a circumferential direction (see Fig. 4, video 1). The second video (Fig. 5) shows a hyperfluorescent spot traveling along a vessel. The distance that the intensely hyperfluorescent spot traveled toward the cornea along a segment of the vessel loop was 43.1 pixels in 0.69 seconds, giving a speed of approximately 0.22 mm/s in that particular segment. 

 

Figure 4

 

Video 1: ICGA scanning laser ophthalmoscope incorporating ART using a 15° lens. Note the segments of pulsatile circumferential flow in the innermost capillary loops.

 

Figure 5

 

Video 2: ICGA scanning laser ophthalmoscope incorporating ART using a 15° lens. Note hyperfluorescent spot ascending toward the cornea and then partially breaking up before descending at a slower rate.

Discussion

The recurrent conjunctival and palisadal vessels and those forming the MCA are derived from the anterior ciliary arteries via the episcleral vessels. ^1–7 Although there have since been several studies using a variety of methods and techniques to image the blood and aqueous vessels of the ocular limbus (reviewed by van der Merwe and Kidson), ¹⁵ there has been little further work on the MCA. Li et al. used optical microangiography to describe the vascular network and suggested that a fraction of the conjunctival plexus arises from terminal vessels that reach the palisades of Vogt to supply the peripheral corneal arcades. ¹⁶ They also noted recurrent vessels in the conjunctival plexus, which run posteriorly to supply the perilimbal area. ¹⁶ The results of this study indicate that the vessels of the corneal periphery extending into the region of the limbus comprising the MCA are a network of vessels consisting of vascular loops with three to four branches, which can be approximated by an elliptical shape with the major axis twice as long as the minor axis. The high tortuosity values and fractal number are also consistent with vascular loops. 

It is not clear whether the angiographic appearances and filling pattern noted in the subjects in this study was truly reflective of a normal healthy eye. Although all of the subjects had unilateral corneal disease, it is possible that unrecognized inflammatory changes altered the blood flow and vessels of the capillary network of the MCA in the fellow eye. In addition, keratitis may also result in nerve damage with cytokine and chemokine gene expression, ^17–19 which may lead to contralateral changes in vascularity. In addition, age and possibly gender may have also contributed to variation in loop size and branching between subjects for similar zone location. Rather than an accurate presentation of MCA parameters in the healthy cornea, we have shown in this study that the MCA can be imaged and measured with the techniques outlined. Clearly further work is needed to characterize the MCA according to location in the cornea and also the disease and characteristics of the subject. 

It was difficult to determine where the vessels of the MCA end and where the limbal vessels began. We were not able to discern a change in vessel parameters across the region (850 μm) included. Although it has been reported that the vessels of the MCA arise from a vessel with the limbal palisade, it was not possible to identify a distinct feeder vessel at the innermost of the five rows. Although further imaging into and across the palisades is needed to determine where the MCA ends and vessels of the palisades begin, it is difficult to analyze this in vivo with the current system as the resolution required to characterize the MCA markedly limits the field of view. 

The innermost row of loops of the MCA, however, appeared to be slightly larger than the average of the four to five rows. It is possible that this, together with an increased path length, leads to a reduced flow velocity of blood flow. In comparison to the size of the perifoveal vascular loops of the retina, the vascular loops of the MCA are bigger. ²⁰ For example, the area contained within a loop of the capillaries of the MCA was 11.87 × 10⁻³ mm² (10.44) compared with 3.95 × 10⁻³ mm² reported in the retinal ganglion cell layer to superficial inner plexiform layer, 5.42 × 10⁻³ mm² in the deep inner plexiform layer to superficial inner nuclear layer, and 6.87 × 10⁻³ mm² in the deep inner nuclear layer. ²⁰  

The pattern of filling, that is, the inferior then superior followed by nasal then temporal, is of interest. It may account for the pattern of deposits seen in various conditions such as the prominence of arcus senilis in the superior and inferior regions. Whether this pattern of filling reflects differences in a vascular supply to these regions or is related, for example, to the greater prominence of palisades in the inferior and superior zones is unclear. The presence, however, of an interlinked vascular network rather than an end artery supply is supported by the circumferential flow of ICG along the inner loops of the MCA evident in the video. Tissue perfusion is proportional to the transit time across a capillary bed, which in turn governs the time available for the exchange of respiratory gases. A change in the speed of blood flow through capillaries is, therefore, an important physiological mechanism to improve oxygen. A problem in measuring blood flow velocity through capillaries by observation, however, is the need for a mark by which blood motion along the vessel can be observed. ²¹ Ivanov used gaps (plasma) between erythrocyte flow to measure velocity. ²¹ Although motion can be seen as evident in the video shown in Figure 5, it is difficult to measure gaps with the use of dyes such as ICG and FA. It was opportune, therefore, to have observed a moving intensely hyperfluorescent spot in the MCA. Although only based on one patient, the average speed of flow in the marginal arcade toward the cornea of 0.22 mm/s, is similar to the velocity of blood flow in the capillaries of the cochlea that has been measured from below 0.1 mm/s to about 0.3 to 0.4 mm/s, with an average velocity of 0.22 mm/s. ²² It can be seen in the second video that the intense hyperfluorescent spot breaks up as the ascending vessel divides and then moves more slowly along the descending limb. In comparison, flow speeds in the retinal parafoveal capillaries are higher, varying from 0.89 ± 0.2 mm/s, ²³ 0.5 and 1 mm/s, ²⁴ 1.82 ± 0.42 mm/s ²⁵ to 3.29 ± 0.45 mm/s, ²⁶ depending on the methods used. Flow rates through capillaries of the cerebral cortex are between 0.3 and 1.5 mm/s, ^20,27–30 and are closer to that observed in the MCA. It is difficult, however, to draw conclusions regarding the function of the MCA from the present data. It would, therefore, be important to investigate the functional activity of the peripheral regions of the cornea and limbus in relation to the vascular filling pattern and speed of flow. Due to the limitation of the resolving power of the retinal camera (Heidelberg Engineering), it is difficult to measure flow in the more proximal limbal and other vessels. Alternative measuring techniques such as laser-Doppler and the newly emerging Doppler optical coherence tomography (OCT) may be used in the future to address this challenge. 

Although we have been able to characterize some of the metrics of the MCA, we have been limited by the resolution needed to measure the diameter of the lumen of the capillaries. In addition, we have made the assumption that the capillary loops of the MCA lie in a flat plane. Improving the resolution of the retinal camera (Heidelberg Engineering) and using its Z-scan acquisition protocol ³¹ may allow measurement of the vessels in different planes, thereby providing volumetrics of the MCA and limbus. In addition, the high resolution and speed of new emerging OCT techniques may have the potential to address this challenge. ³² Studies are needed to further characterize the MCA according to location. In particular, from a clinical application, it would be useful to follow patients longitudinally to determine how the morphology of the MCA changes in different diseases and from which vessel CoNV arise, similar to the process that occurs in diabetic retinopathy. This may improve our understanding of the cornea's response to disease and permit the development and refinement of treatments aimed at modifying the vascular response to injury. 

Acknowledgments

Disclosure: Y. Zheng, None; A.E. Kaye, None; A. Boker, None; R.K. Stewart, None; A. Tey, None; S. Ahmad, None; C.E. Willoughby, None; A.J. Bron, None; S.B. Kaye, None