Imaging and Evaluation of Corneal Vascularization Using Fluorescein and Indocyanine Green Angiography

Abstract

Purpose.: To evaluate indocyanine green angiography (ICGA) and fluorescein angiography (FA) in imaging and quantifying corneal neovascularization (CNV).

Methods.: Patients with CNV were studied using a standardized protocol of color digital photography, FA, and ICGA. Images were graded independently by two observers and assessed for quality, phases of fluorescence, and leakage. Areas of CNV and vasculature geometric properties were analyzed and quantified by an automated program.

Results.: Twenty-three patients with good quality images were included. Mean times to appearance of ICG and fluorescein were 17 and 20 seconds (P = 0.10). Best images for analysis were obtained at 64 seconds for ICGA and 47 seconds for FA. CNV not apparent on color or FA, particularly in the presence of scarring, was well delineated by ICGA. Leakage of ICGA did not occur. Fluorescein leakage from apical CNV images occurred significantly earlier (32 seconds) in patients with CNV of <6-month duration than those of >1-year (50 seconds) duration (P = 0.04). Mean area of CNV and vessel diameter were similar with ICGA (8.79 mm², 0.058 mm) or FA (7.74 mm², 0.054 mm) but significantly larger than on color (1.94 mm², 0.026 mm) images (P < 0.01). Vessel tortuosity was similar on ICGA (1.16), FA (1.17), and color (1.15) (P = 0.27).

Conclusions.: Combined use of FA and ICGA are valuable tools with which to assess CNV and provide better vessel delineation than can be obtained with only color images. Parameters used to assess CNV, such as leakage, area, diameter, and tortuosity, may be useful measures for evaluating treatment. Videography is useful for detecting early leakage.

In its healthy state, the cornea is devoid of blood or lymphatic vessels. ^{1 –3} Corneal neovascularization (CNV) is a common response to a variety of insults, such as trauma, chemical injury, inflammation, and infection. ^{3 –5}  

It has been estimated that in the United States, 1.4 million patients per year develop CNV; 12% of them also experience a decrease in visual acuity. ⁶ In addition, the development of CNV results in the loss of the immunoprivileged status of cornea ⁷ and is a strong risk factor for immune rejection after corneal transplantation. ^8,9 Several angiogenic growth factors, such as vascular endothelial growth factor (VEGF), have been implicated in corneal angiogenesis. The identification of naturally occurring antiangiogenic factors together with the development of inhibitors of angiogenesis has led to several potential treatment modalities to suppress and treat CNV. ^{10 –15}  

A key requirement for the evaluation of any potential treatment, however, is the ability to quantitatively compare CNV before and after any intervention. The most common method for quantifying CNV is by analyzing photographic images of the cornea usually taken during slit lamp biomicroscopy. Lin et al. ¹⁰ used this method to evaluate murine models of corneal vascularization. Similarly, Sharma et al. ¹³ and Dasterdji et al. ¹⁴ investigated the effect of bevacizumab eye drops on CNV by identifying changes in neovascular area, invasion area, and vessel caliber using graphics editing software. Unfortunately, not all blood vessels are visible, even on the best quality color images, especially in the presence of scarring. In addition, it is very difficult to image very small blood vessels using this method or to determine vessel maturity or function. 

These are not, however, new issues. Although fluorescein angiography (FA) was recognized as a useful tool for the evaluation of CNV in 1971, ¹⁶ there has been little subsequent interest in its use. This might have been a reflection of the available technology and imaging systems. Similarly, there has been little interest in the use of indocyanine green (ICG) for the evaluation of CNV. ICG is a water-soluble tricarbocyanine dye with a peak spectral absorption at 800 to 810 nm ¹⁷ that has proved particularly useful for imaging the choroidal and retinal vasculature. After intravenous administration, ICG is rapidly bound to plasma protein and, because of the size of the resultant molecules, does not leak as much as fluorescein dye from blood vessels, allowing detailed imaging of the vasculature. Combining FA and ICG angiography (ICGA), therefore, offers the potential to identify and to assess the maturity of CNV. One case has been reported in which ICGA and FA were used in one patient to evaluate the effect of photodynamic therapy on CNV. ¹⁸ With the advent of new imaging systems and analytical processes, there is an opportunity to evaluate the role of ICGA and FA to image, stage, and quantify CNV. 

Patients and Methods

Consecutive patients older than 18 years with at least 3 mm of clinically evident CNV on slit lamp biomicroscopy because of various pathologies were included. If a patient had bilateral CNV, the eye with the greater CNV was selected for imaging. Patients were excluded if they had a history of allergy to fluorescein, iodides, and shellfish. After inclusion, patients were divided into those with clinically documented CNV of <6-month and of >1-year duration. Informed consent was obtained and the study was conducted according to the tenets of the Declaration of Helsinki. 

Color Images

Color images of the cornea were recorded using a slit lamp mounted digital system (SL-D Digital Slit Lamp; Topcon, Tokyo, Japan). The entire cornea was imaged from limbus to limbus using 10× magnification of the region of interest (ROI). Illumination was from a 45° angled beam on slit lamp biomicroscopy with a diffuser filter and a variable flash intensity. An anterior-posterior image with diffuse beam and no filter was also obtained. Images with green filter and open beam at magnifications of 10× were obtained. A total of 18 color images were obtained for each patient. 

Indocyanine Green and Fluorescein Angiography

Both ICGA and FA were acquired using a scanning laser ophthalmoscope (HRA2; Heidelberg Engineering, Heidelberg, Germany). Five milliliters of 5 mg/mL indocyanine green dye (Pulsion Medical Systems, Munich, Germany) was injected into a peripheral arm followed immediately by videography for 25 seconds. Single-frame ICGA photographs of the whole cornea capturing corneal blood vessel fluorescence every 3 to 5 seconds were taken for 3 minutes in high-resolution mode incorporating automatic real-time (ART) software. This was followed by an intravenous injection of 3 mL of 20% sodium fluorescein (Martindale Pharmaceuticals, Essex, UK). The photographs were taken up to 3 minutes, similar to ICGA. Late ICGA and FA images were taken at 5 and 10 minutes. 

Image Analysis

Video pictures taken immediately after the injection of fluorescein or ICG were analyzed independently by two observers (HS and AT) for appearance of dye. Color images were analyzed for quality and clarity of ROI. The best ICGA, FA, and color images were independently selected by same two observers (HS and AT) based on the following qualitative subjective parameters. Quality grading for ICGA and FA was categorized as 0 to 4 (0, no vessel discernible; 1, poor vessel delineation; 2, good vessel delineation; 3, very good vessel delineation; 4, excellent vessel delineation). The images with good vessel delineation would have at least 50% of vessels clearly evident with distinct boundaries and hyperfluorescence. Quality grading for color images was categorized as 0 to 3 (0, no focus on ROI; 1, poor focus on ROI, no details of corneal scar/vessel; 2, acceptable focus on ROI, identifiable corneal scar/vessel; 3, very good or excellent focus on ROI, details of corneal scar and vessels evident). Only patients with images of quality score 3 or more for both ICGA and FA and quality score 2 or more for color were included for further analysis. Haziness and increasing fluorescence of corneal stroma adjacent to the blood vessels on FA images was considered as evidence of leakage. The relationship between leakage of fluorescein dye and duration of CNV was also examined. 

The best available ICGA, FA, and color image of the study eye of each patient was exported in TIFF format for the purpose of quantitative analysis. The area of CNV and geometric properties of the vessels were determined on the selected images using an in-house automated program written in numerical computing language (MatLab R14; The MathWorks Inc., Natick, MA). This objective analysis was carried out on a computer (configurations: Windows XP Service Pack 2 [Microsoft Corporation, Redmond, WA], Intel Core 2 [Intel Corporation, Mountain View, CA], 2.66 GHz, and 3.25 GB of RAM). Similar to our previous strategy in assessing a foveal avascular zone in FA images, ¹⁹ this semiautomated program consists of the following process steps. The first step involved the identification of limbus manually on the image to measure the corneal diameter in pixels. This was followed by estimation of pixel resolution (mm/pixel), defined as the ratio between the diameter of the cornea and the number of pixels. A sub-image containing all the corneal vessels was defined by hand and was enhanced through a Gaussian filter to remove noise, followed by the application of selective enhancement filters initially described by Li et al. ²⁰ to the smoothed image to enhance all the potential vessels as linear structures. This filter was adopted for its simplicity and effectiveness. The enhanced image was then converted to a binary image in which all the pixels with values higher than a predefined value were marked as 1 (vessel pixels), the rest as 0 (background pixels). The binary image was further cleansed by removing objects smaller than 10 pixels. The area of CNV (mm²) was defined according to the number of pixels and the resolution of pixels in x and y axes. After segmentation of the corneal vessels, a three-step automatic analysis process was adopted to quantify the geometric property of the vasculature tree structure by adapting a well-established semiautomatic method described by Martinez-Perez et al. ²¹ First, the centerlines of the segmented vessels were determined by a mathematical morphologic thinning operation. The significance points (branch points and terminal points) were then identified and used to segment the vascular tree into individual segments. For each vessel segment, its geometric features (length, area, diameter along the path, average diameter, and tortuosity) were measured and used to describe the overall properties of each patient's CNV. Tortuosity was defined as the length to cord ratio of the vessel segment. ²¹ Two observers (AT and YZ) ran the program on all the selected FA and ICGA images in a masked pattern, and one observer (AT) repeated the test on the same day. 

Statistical Analysis

Levels of agreement for categorical data were tested using Fleiss' kappa statistic (κ). Interpretation of levels of agreement was based on that described for two (binary) categories for each patient. A value of <0.2, κ was considered slight, 0.2 < κ < 0.4 fair, 0.4 < κ < 0.6 moderate, 0.6 < κ < 0.8 substantial, and κ > 0.8 as almost perfect agreement. ²² For continuous data, intraclass correlation coefficients (ICCs) were used to evaluate repeatability and reproducibility. Student's t-test, Mann-Whitney U test, and ANOVA were used to test for a difference between means. Minitab version 16 (Minitab Inc., State College, PA) was used for the statistical analysis. P ≤ 0.05 was considered statistically significant. 

Results

Patient Characteristics

Thirty-nine patients with CNV underwent FA, ICGA, and color photography. Seventy-two percent of ICG and 67% of the FA images were grade 3 or higher. Eighty-two percent of color images were grade 2 or higher. Images of patients with both ICG and FA of grade 3 or higher and color of grade 2 or more were used for analysis. Sixteen patients were excluded from further analysis because the quality of their FA, ICGA, or color images did not meet these criteria. Of the 23 patients included, there were 11 men and 12 women ranging in age from 19 years to 77 years (Table 1). CNV was associated with presumed or microbiologically confirmed herpes simplex keratitis (HSK) in 14 patients, ocular surface disease in three patients, bacterial keratitis in two patients, and acanthamoeba keratitis, marginal keratitis, post penetrating keratoplasty (PK), and post deep anterior lamellar keratoplasty in one patient each (Table 1). All patients except six had inactive keratitis. The onset of the duration of CNV ranged from 1 week to more than 10 years (Table 1). 

Table 1.

 

Patient Characteristics

Patient	Age (y)	Diagnosis*	Duration of CNV	Scar†	Keratitis‡	Treatment§	Time (s)¶
							ICG	FA	Leakage
1	77	OSD	31 y	S	No	N	13	19	31
2	59	HSK	5 mo	S + M	No	Av	10	8	19
3	45	HSK	8 y	M + D	No	St, Av	34	29	37
4	66	HSK	5 y	S	No	St	NA	24	39
5	61	OSD	6 y	S	No	N	NA	19	49
6	34	HSK	4 mo	S	No	St	11	13	30
7	19	HSK	13 mo	S	No	St	21	19	54
8	25	AK	5 mo	M + D	No	St	22	41	53
9	29	HSK	6 y	S	No	N	NA	28	118
10	64	HSK	7 y	M + D	No	St	NA	27	90
11	39	OSD	19 mo	S	No	St	16	20	51
12	56	HSK	11 y	S + M	No	St	18	18	36
13	32	HSK	14 mo	S	No	N	24	27	53
14	43	BK	3 wk	S	Yes	St, Am	10	10	22
15	28	DALK	4 mo	I	No	St	21	NA	44
16	68	HSK	5 mo	S	No	N	16	21	39
17	60	HSK	2.5 y	S + M	No	N	15	11	23
18	43	PK,	3 wk	S	Yes	St	14	18	36
19	67	HSK	1 wk	S	Yes	Av, Am	13	16	23
20	68	HSK	1 wk	S	Yes	St, Am	15	16	29
21	31	BK	1 wk	S	Yes	Am	15	17	25
22	54	BK	2 y	S	Yes	N	19	22	41
23	69	HSK	13 mo	S + M	No	N	20	19	30
Mean	—	—	—	—	—	—	17	20	42
SD	—	—	—	—	—	—	6	7	23

*  Diagnosis: OSD, ocular surface disease; HSK, herpes simplex keratitis; AK, acanthamoeba keratitis; BK, bacterial keratitis; DALK, deep anterior lamellar keratoplasty; PK, penetrating keratoplasty.

†  Stromal scarring: S, superficial; M, med; D, deep; I, interface.

‡  Keratitis: clinically active keratitis based on presence of inflammation, infiltrate, ulceration, or stromal edema.

§  Treatment (topical at the time of angiography): N, none; Av, antiviral; St, steroid; Am, antimicrobial.

¶  Time (in seconds) to first appearance of ICG, FA, and leakage of fluorescein: NA, not available due to poor quality of initial video frames.

Analysis of ICGA, FA, and Color Images

Interobserver agreement for grading the quality of image assessment was substantial for both ICGA (κ = 0.65; P < 0.01) and FA (κ = 0.77; P < 0.01) but moderate for color images (κ = 0.51; P < 0.01). There were no significant differences within or between observers for the measurement of area (P = 0.75 and P = 0.82), tortuosity (P = 0.43 and P = 0.69), and diameter (P = 0.74 and P = 0.32), with consistent intraobserver and interobserver agreement for measurement of total vessel area (ICC, 0.97 and 0.93), tortuosity (ICC, 0.84 and 0.77), and diameter (ICC, 0.93 and 0.95), respectively (P < 0.01). 

Mean times to appearance of ICG and fluorescein were 17 seconds (SD, 6 seconds) and 20 seconds (SD, 7 seconds) (P = 0.10) (Table 1). There was no difference between the time to appearance of ICG or fluorescein in patients with CNV of <6 months (16 seconds [SD, 5 seconds] and 19 seconds [SD, 10 seconds]; P = 0.36) or more than 1 year in duration (19 seconds [SD, 7 seconds] and 21 seconds [SD, 5 seconds]; P = 0.73). Some of the video pictures, however, were of poor quality or were not available; hence, the appearance of fluorescence (ICGA or FA) was difficult to determine in five patients (Table 1). Best images were obtained at 64 seconds (SD, 41 seconds) for ICGA and 47 seconds (SD, 19 seconds) for FA (P = 0.01). Details of CNV, in particular those that are imperceptible on color, were well delineated with ICGA (Figs. 1, 2). Even though CNV was highlighted with FA, the quality of fluorescence was partially blocked in the presence of scarring. The degree and depth of scarring also appeared to affect the degree of blocking of the fluorescence (Figs. 1, 2). 

Figure 1.

 

Corneal neovascularization in acanthamoeba keratitis (patient 8). Color image (a), fluorescein angiography at 63 seconds (b), indocyanine green angiography at 48.5 seconds (c), and digital image analysis of ICGA (d).

Figure 2.

 

Corneal neovascularization in presumed herpes simplex keratitis (patient 1). Color image (a), fluorescein angiography at 39 seconds (b), indocyanine green angiography at 56 seconds (c), and digital image analysis of ICGA (d).

Area of CNV and Vessel Parameters

CNV area (Figs. 1d, 2d) was measured on the best ICGA and FA images using the method and software program described. Although there was no significant difference in the mean area of CNV using ICGA (8.79 mm² [SD, 6.12 mm²]) and FA (7.74 mm² [SD, 5.11 mm²]) (P = 0.15), it was significantly smaller using color images (1.94 mm² [SD, 1.18 mm²]) (Table 2). Parameters of the blood vessels (diameter and tortuosity) within each patient's ROI measured using ICG, FA, and color are shown in Table 3. The overall mean and median diameter (mm) of blood vessels for all patients were similar using ICGA (0.058 mm, 0.055 mm [SD, 0.014 mm]) and FA (0.054 mm, 0.052 mm [SD, 0.017 mm]) (P = 0.15) but larger than measured using color images (0.026 mm, 0.026 mm [SD 0.004 mm]) (P < 0.01). Tortuosity of the blood vessels (mean, median, SD) was, however, similar for ICGA (1.16, 1.08 [SD, 0.03]), FA (1.17, 1.08 [SD, 0.02]), and color (1.15, 1.08 [SD, 0.02]) (P = 0.27). There was a positive skew for the distribution of vessel tortuosity for ICGA (4.74), FA (5.00), and color (5.21), indicating the presence of a few particularly tortuous vessels within the ROI. The value of these measures of vessel characteristics is apparent in patient 18, who developed active keratitis with CNV after previous PK for HSK. Over a 4-week period, there was a significant reduction in the area of CNV after treatment with an antiviral and topical steroid, from 3.32 mm² to 2.44 mm², accompanied by a reduction in mean vessel diameter (0.042–0.030 mm) and tortuosity (1.23–1.19) (Figs. 3, 4). 

Table 2.

 

CNV Area Measured from the Best Selected ICGA, FA, and Color Images

Patient	ICGA	FA	Color
1	10.66	14.06	4.04
2	6.18	8.46	2.10
3	10.41	11.45	2.00
4	11.26	3.89	1.99
5	26.78	17.67	3.86
6	4.33	4.98	NA
7	3.65	4.67	0.92
8	9.01	7.89	3.36
9	3.15	2.59	0.82
10	17.94	11.38	1.28
11	1.19	1.36	1.10
12	12.74	9.55	2.65
13	3.77	4.38	NA
14	3.44	3.64	1.16
15	5.69	4.96	0.43
16	2.84	3.09	1.35
17	18.27	20.25	0.80
18	4.72	5.31	0.33
19	9.28	4.89	1.20
20	10.96	12.65	3.13
21	12.23	12.75	3.39
22	9.84	3.59	3.46
23	3.94	4.48	1.46
Mean	8.79	7.74	1.94
SD	6.12	5.11	1.18

 All values are in mm². NA, not available because of the reduced quality of the color images from corneal scars.

Table 3.

 

Vessel Parameters of CNV within the ROI: Diameter and Tortuosity

Patient	Vessel Diameter, mm mean (median) SD			Vessel Tortuosity mean
	ICG	FA	Color	ICG	FA	Color
1	0.066 (0.031)	0.066 (0.041)	0.029 (0.019)	1.14	1.20	1.17
	0.060	0.051	0.025
2	0.078 (0.026)	0.068 (0.032)	0.024 (0.009)	1.14	1.15	1.14
	0.071	0.063	0.024
3	0.067 (0.032)	0.085 (0.043)	0.023 (0.010)	1.12	1.16	1.18
	0.053	0.083	0.025
4	0.077 (0.045)	0.061 (0.031)	0.029 (0.012)	1.14	1.18	1.16
	0.080	0.050	0.027
5	0.063 (0.030)	0.051 (0.019)	0.031 (0.015)	1.15	1.15	1.15
	0.065	0.061	0.030
6	0.072 (0.018)	0.080 (0.046)	NA	1.12	1.19	NA
	0.083	0.082
7	0.023 (0.009)	0.022 (0.010)	0.019 (0.008)	1.14	1.15	1.18
	0.019	0.020	0.017
8	0.050 (0.023)	0.046 (0.016)	0.025 (0.010)	1.17	1.18	1.15
	0.053	0.051	0.027
9	0.075 (0.048)	0.072 (0.058)	0.027 (0.019)	1.19	1.17	1.16
	0.056	0.059	0.026
10	0.058 (0.034)	0.050 (0.023)	0.022 (0.009)	1.14	1.20	1.15
	0.057	0.050	0.026
11	0.040 (0.021)	0.036 (0.020)	0.029 (0.010)	1.18	1.17	1.14
	0.035	0.032	0.027
12	0.057 (0.020)	0.063 (0.019)	0.021 (0.009)	1.16	1.16	1.15
	0.064	0.069	0.018
13	0.062 (0.030)	0.069 (0.047)	NA	1.18	1.18	NA
	0.060	0.050
14	0.058 (0.019)	0.070 (0.042)	0.028 (0.013)	1.15	1.22	1.15
	0.069	0.071	0.027
15	0.051 (0.035)	0.047 (0.016)	0.024 (0.007)	1.18	1.13	1.14
	0.045	0.049	0.024
16	0.061 (0.029)	0.055 (0.028)	0.035 (0.023)	1.21	1.22	1.14
	0.053	0.046	0.029
17	0.061 (0.034)	0.049 (0.021)	0.019 (0.009)	1.15	1.16	1.17
	0.046	0.044	0.017
18	0.039 (0.022)	0.049 (0.018)	0.033 (0.009)	1.20	1.20	1.11
	0.034	0.047	0.035
19	0.061 (0.038)	0.044 (0.040)	0.025 (0.011)	1.15	1.16	1.12
	0.048	0.039	0.027
20	0.041 (0.018)	0.036 (0.016)	0.031 (0.013)	1.19	1.17	1.16
	0.038	0.036	0.029
21	0.073 (0.034)	0.068 (0.027)	0.026 (0.011)	1.16	1.19	1.15
	0.071	0.072	0.026
22	0.065 (0.036)	0.027 (0.011)	0.024 (0.012)	1.21	1.17	1.16
	0.063	0.029	0.024
23	0.034 (0.019)	0.035 (0.016)	0.027 (0.017)	1.15	1.16	1.16
	0.035	0.037	0.025
Mean	0.058	0.054	0.026	1.16	1.17	1.15
SD	0.014	0.017	0.004	0.03	0.02	0.02

 NA, not available.

Figure 3.

 

Demonstration of vessel segmentation and analysis of geometric features at entry and during follow-up after treatment (patient 18). ROI (a), segmentation of vessels (b), terminal points and centerline of vessels (c). ROI at 4 weeks of follow-up (d). Vessel segmentation (e) and terminal points and centerline of vessels (f).

Figure 4.

 

Frequency distribution of vessel segments in terms of their diameters over the ROI at entry and at follow-up at 4 weeks, corresponding to Figure 3. Note the reduction in vessel diameters at follow-up.

Leakage of Fluorescein and ICG

Although vessel architecture was very well delineated by ICG, there was no leakage. FA was useful to detect apical leakage (Figs. 5a, 5b; Table 1). The mean leakage time of fluorescein dye from CNV was 42 seconds (SD, 23 seconds), with leakage occurring in all patients by 5 minutes. Leakage of fluorescein from CNV was significantly earlier (32 seconds) in patients with a history of CNV of <6-month compared with those of >1-year (50 seconds) duration (P = 0.04). Leakage of fluorescein is apparent in patient 6, in whom the CNV was caused by presumed HSK of 4 months' duration and in patient 12 because of HSK of 11 years' duration. Apical hyperfluorescence caused by leakage of dye is seen first at 30 seconds (Fig. 5b) and 36 seconds (Fig. 5d). Progression of leakage at 49 seconds and at 5 minutes is shown for patient 12 (Figs. 5e, 5f). In contrast, no leakage is apparent with ICGA (Figs. 5a, 5c). 

Figure 5.

 

Indocyanine green and fluorescein angiographic images of patients 6 (a, b) and 12 (c–f). The commencement of apical leakage of fluorescein is seen at 30 and 36 seconds (b, d). Progression of leakage is shown in patient 12 at 49 seconds and at 5 minutes (e, f). In contrast, no leakage is apparent with indocyanine green (a, c).

Fluorescence from Topical Fluorescein

Some of the initial patients received, at the commencement of angiography, topical fluorescein before intravenous fluorescein. The residuum of fluorescein and uptake within the corneal stroma affected the quality of FA images of CNV, as evident in Figure 6a (patient 10). This shows both a fluorescein level in the tear meniscus partially obscuring the lower part of the cornea and uptake into the corneal stroma. This made it difficult to determine the degree of leakage from the stromal vessels. Repeat FA showed no stromal fluorescence in the absence of topical fluorescence (Fig. 6b). 

Figure 6.

 

Fluorescein angiographic image of CNV (patient 10) at the start of imaging in the presence of superficial to deep stromal scarring exhibiting fluorescence in the presence of topical fluorescein (a). Fluorescein image in the absence of topical fluorescein (b) with no stromal staining and more evident apical leakage at 90 seconds.

Discussion

Angiogenic corneal insults result in a cascade of inflammatory reactions that result in the release of angiogenic agents that promote CNV formation. Although the resultant neovascularization is a defense mechanism to the inciting cause, there is often associated scarring, edema, lipid deposition, and persistent inflammation that may significantly reduce visual acuity. Not all CNV is clinically evident; 20% of corneal specimens obtained during corneal transplantation show histopathologic evidence of CNV. ²³  

Areas of CNV have been evaluated using color images in recent studies. ^13,14 Sharma et al. ¹⁴ evaluated the role of angiotensin converting enzyme on CNV artificially induced in New Zealand rabbits. The CNV area was manually marked and measured using image editing software (Photoshop CS2; Adobe Systems, San Jose, CA) and ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html). Dasterdji et al. ¹³ used Adobe Photoshop CS2 and a MatLab program to quantitatively evaluate the area of CNV on slit lamp digital images. The difficulty with this approach for the quantification of the CNV area is that not all blood vessels are visible even on the best quality color images, especially in the presence of scarring, as evident in patients 1, 8, and 10 in our study. 

To better assess CNV, Easty and Bron ¹⁶ obtained FA images of CNV in 250 patients. They observed no definitive pattern of abnormal vasculature of the cornea and found it difficult to differentiate arterial and venous phases because the flow is rapid compounded by the presence of multiple vascular beds in the same cornea. Leakage into corneal stroma was, however, noted in the presence of inflammation and vessel immaturity. ¹⁶ Easty and Bron ¹⁶ also described the phenomenon of pseudofluorescence and apical leakage; the latter was particularly evident in our case series. FA for the evaluation of CNV was, however, neither developed further nor commonly used, possibly because of difficulty in obtaining good quality images without the access of confocal laser ophthalmoscopy and because of the absence of automated methods for quantifying CNV and the absence of an effective treatment. 

ICG, which is larger and more protein bound than fluorescein, is retained in the blood vessels, providing good vessel definition. In addition, ICGA, with its near infrared spectrum of absorbance, appears to be able to define CNV more effectively than color or FA images, particularly blood vessels located more deeply in the cornea, beneath corneal scarring. Although not measured in this study, ICGA also allows stereoscopic images to evaluate the depth of CNV, as occurs with imaging of retinal and choroidal vasculature. For example, we can see that in patient 8, mid to deep stromal scarring blocks the identification of blood vessels (Fig. 1). In contrast, ICGA allowed a clear delineation of corneal blood vessels compared with the color image. In comparison to ICGA, estimation of the area of CNV from FA images is difficult because of leakage and staining of the corneal stroma. For quantitative analysis, we have applied a linear enhancement filter to highlight vessels that result in their automatic detection. To some degree, this alleviates a major challenge in the quantitative CNV analysis of color images attributed to poor image quality resulting from the low contrast of vessels. With this technique, both FA and ICG provided similar measurements of the CNV area, which was larger than that detected on color images. This difference in area may reflect that FA and ICGA, in particular, identified and delineated vessels that were not apparent clinically or on color images. In terms of vessel parameters, ICGA and FA and, to a lesser extent, the color images allowed a good measure of vessel diameter and tortuosity across each patient's ROI. This may be particularly useful for evaluating the natural history and response to the treatment of CNV, as in the example shown in Figures 3 and 4. Although mean vessel diameter was significantly smaller on color than on FA and ICGA, there was no significant difference in vessel tortuosity. It is not clear whether the difference in vessel diameter between color and ICGA or FA is an indication of vessel wall thickness. This will require further study, possibly combined with histology and confocal microscopy. 

Although the quality of corneal images obtained with FA, particularly in the presence of scarring, was inferior to that obtained with ICGA, the leakage of fluorescein may reflect vessel maturity, and it offers the potential to functionally stage CNV. Although late leakage at 5 minutes was observed in all patients, similar to the findings of Easty and Bron, ¹⁶ leakage began at earlier time points. In particular, the leakage of fluorescein from the leading apical vessels occurred significantly earlier in patients with CNV of <6-month (32 seconds) than in those with CNV of >1-year (50 seconds) duration. If early leakage is indicative of vessel maturity, it may be useful in assessing response to treatment; longitudinal studies are needed with regard to this. It is important, however, that FA be undertaken in the absence of topical fluorescein. The entry of topical fluorescein into the corneal stroma, as evident in images recorded in the presence of fluorescein filters, may provide useful information regarding the epithelial layer, but this must be undertaken separately from FA. The absence of leakage of ICG enables the quantitative analysis of the blood vessels in which the scarring and leakage of fluorescein impairs the ability to capture the full extent of CNV with FA. 

The combined use of FA and ICGA is valuable in assessing CNV and appears to provide better vessel delineation than that obtained with only color images. The parameters used to assess CNV—such as leakage, area, diameter, tortuosity, and other measurable characteristics, such as branching and segment length—are likely to be useful measures for evaluating response to treatment. The best images for analysis were obtained at approximately 1 minute for ICGA and slightly earlier (47 seconds) for FA. Videoangiography is particularly useful for detecting early leakage with FA; recording should begin at approximately 15 seconds to allow detection of the appearance of fluorescein and subsequent leakage. 

There was a learning curve in the imaging process. Approximately 25% of patients had to be excluded because of poor quality of either FA or ICGA images, and up to 40% were excluded based on our inclusion criteria, primarily because of focusing on the earlier and more prominent fluorescence in iris vessels. This was a learning phenomenon; technicians should be trained in this type of imaging if good quality and repeatable images are to be obtained for comparative quantitative analysis. Although no adverse reactions to FA and ICGA were encountered, it is still an invasive investigation with potential anaphylaxis and other side effects. 

Footnotes

 Disclosure: D.R. Anijeet, None; Y. Zheng, None; A. Tey, None; M. Hodson, None; H. Sueke, None; S.B. Kaye, None

The authors thank Stephen Pearson, Jerry Sharp, William Hooley, and Gillian Lewis (Ophthalmic Imaging Department, St. Paul's Eye Unit, Liverpool, UK) for their roles in performing the angiograms.