Fundus Photography for Measurement of Macular Pigment Density Distribution in Children

Abstract

purpose. To evaluate a photographic procedure for reflectometry of the topographic distribution of macular pigment density in normal pediatric subjects.

methods. Digitized blue (480 nm) and green (540 nm) photographic images were aligned and subtracted to generate optical density difference maps. An 8° × 8° area concentric with the fovea was analyzed. Gaussian curves were fitted through the foveola along the vertical and horizontal meridians. The peak density and full widths at half maximum (FWHMs) were calculated. The subjects (n = 23; median age 10.5 years) had normal eyes and good acuity.

results. The peak macular pigment (MP) density was 0.13 ± 0.04 density units (DU) which is at the lower end of the range previously obtained by other reflectometry procedures. Density distributions were circularly symmetrical. The FWHM ranged was 2.4° ± 0.5°. Neither MP nor FWHM varied significantly with age.

conclusions. The photographic method is feasible and provides quantitative assessment of topographic properties of macular pigment in young subjects. Future application to clinical studies of pediatric patients is envisioned.

At the center of the macula, the fovea has a protracted course of development ^{1 2} and constrains the development of spatial vision and visual acuity. ³ Disorders of the pediatric macula, such as maculopathies, ⁴ variants of albinism, ⁵ incontinentia pigmenti, ^{6 7} and retinopathy of prematurity ^{8 9} are incompletely evaluated by direct or indirect ophthalmoscopy. Imaging procedures designed for cooperative adults have limited application in pediatric patients. Thus, the quest for a practicable, quantitative approach to evaluation of the pediatric macula, for both research and clinical application, persists. Herein, a photographic technique is described that quantitatively evaluates the topographic density distribution of macular pigment (MP) in pediatric subjects. The yellow MP is found mainly in the axons of the foveal cones. ^{10 11} These radially arranged axons, the fibers of Henle, are densely packed near the fovea and become less densely packed with increasing eccentricity. The MP topography represents primarily the axonal structure of the central macula, and thus is a proxy for the structure of the center of the macula. 

Measurement of peak MP density is derived from a difference between near peak (480 nm) and near zero (540 nm) absorbency. Psychophysical, fundus autofluorescence, and fundus reflectance techniques have been used to measure peak MP density. Psychophysical procedures yield a range of peak MP densities of 0.0 to 1.0 density units (DU). ^{12 13 14 15 16 17} By the autofluorescence method, ^{18 19} in which the incident light is absorbed by the MP before exciting the pigment epithelial lipofuscin fluorescence, the peak MP densities are 0.1 to 1.0 DU. Compared with psychophysical and fluorescence methods, the reflectometry procedures yield lower peak MP densities. (0–0.45 DU). ^{19 20 21 22 23}  

In the present study, photography with a conventional fundus camera is used to study MP topography in pediatric subjects with normal fundi. The method is based on fundus reflectometry and involves a two-wavelength procedure. ^{24 25 26 27} Photographs are taken using blue (480 nm) and green (540 nm) illumination. In large part, the light incident on the fovea traverses the MP twice with reflection at layers posterior to the MP including the photoreceptors, the retinal pigment epithelium, and the choroid. ^{15 16 17 18 19 28} The MP absorbs blue light more than green light. ^{29 30 31 32} Subtraction of the aligned green and blue images, and appropriate scaling of the density difference, yields the topographic distribution of the MP density. Herein, both the peak MP density and the topography of MP density were determined. A preliminary report of this work has been published in abstract form. ³³  

Methods

Subjects

Twenty-three subjects, aged 6 to 20 years (median, 10.5 years), participated. All had normal ocular structures and Snellen acuities of 20/20 or better. None had high refractive errors. The median spherical equivalent was plano (range, −3.00 to +1.50 D). Written, informed consent from the parents and assent of the participants were obtained. The study conformed to the tenets of the Declaration of Helsinki and was approved by the Children’s Hospital Committee on Clinical Investigation. 

Fundus Photography

Pupils were dilated with cyclopentolate 1% and phenylephrine 2.5%. Photographs of a 30° retinal field, including the optic disc and macula, were obtained with a fundus camera (NFC-50; Nikon, Tokyo, Japan; TMY 135-24 film; Eastman Kodak, Rochester, NY). In 10 of the 23 subjects, photographs of both the right and left maculas were taken. Interference filters (Ditric; Corion Corp., Holliston, MA; peak transmission 480 and 540 nm, full width at half maximum [FWHM], 25 nm) were placed alternately in the camera slot that ordinarily accommodates the manufacturer’s glass filter for red-free photographs. Alignment and focusing of the fundus were done using green light (540 nm filter) typically for a duration of 4 to 5 minutes; retinal irradiance was 140 to 250 μW/cm² (5.6–5.9 log photopic trolands). Such an exposure bleaches 75% to 85% of rhodopsin and 93% to 96% of the cone photopigments. Radiant exposures for the photographic flash were 3.7 and 4.3 mJ/cm² for the 540- and 480-nm exposures which were 3 to 6 ms in duration. All these light levels are below the maximum permissible exposures according to the American National Standards Institute (ANSI) light safety standard. ³⁴ To calibrate the characteristics of the film negative (relationship between film transmission and exposure), photographs of a calibrated neutral density wedge also were taken on the same film at each session. 

Algorithm

The macular pigment density D(x,y)₄₆₀ is calculated by ¹⁹  

\[D(x,y)_{460}{=}0.5\ {\cdot}\ {\{}log\ (E\mathrm{(ref)}_{480}/E\mathrm{(ref)}_{540}){-}\mathrm{log}\ (E(x,y)_{480}/E(x,y)_{540}){\}}/(K_{480}{-}K_{540})\]

where E(x,y) is the film exposure at any given retinal location (x,y) at the indicated wavelength and K ₄₈₀ and K ₅₄₀ are the known extinction coefficients of the MP at 480 and 540 nm, relative to that at 460 nm. ²⁹ The exposure E(ref) is derived from the mean film exposure at reference sites 4° eccentric. The algorithm assumes that all the incident light traverses the MP twice and that the spectrum of reflective layers posterior to the MP at the fovea is proportional to that at the reference sites. ¹⁹  

Analysis of Photographic Images

Negatives of the 480- and 540-nm images (Fig. 1) and of the calibrated wedge were scanned and digitized (Scanjet, 6100 C; Hewlett-Packard, Palo Alto, CA) with a resolution of 600 dpi. Transmissions of the negatives of the blue and green images (units: 0–255 gray levels) were converted into film exposures, log E, using a calibration curve (fifth degree polynomial fit through the data of the calibrated wedge). The same calibration curve was used for the 480- and 540-nm images. 

First coarse, then fine, alignment of the 480- and 540-nm log E maps was performed. For coarse alignment, clearly recognizable structures within the vicinity of the macula, such as bifurcation of vessels, were interactively marked. The marked structures were then used to bring the 480- and 540-nm images into approximate coincidence by lateral displacement. The 540- and 480-nm images were subsequently smoothed by a seven-pixel spatial low-pass filter. For fine alignment, cross-correlation of an area approximately 100 × 100 pixels was calculated around the landmark (0,0). This cross-correlation shifted the 480-nm image with respect to the 540-nm image ± 20 pixels. Optimal alignment was defined as the position at which the cross-correlation reached its maximum value. Fine-alignment estimates differed by 0 to 10 pixels from that of the coarse alignment. 

Retinal distances in pixels were converted to degree of visual angle by assuming that the diameter of the optic disc was constant across all subjects and equal to 1.77 mm ³⁵ and that all were emmetropic with a posterior focal length of 22.4 mm. ³⁶ Neglecting interindividual differences in refractive error results in an error of less than 5% in magnification. Horizontal and vertical diameters of the optic disc, in pixels, were measured on the 540- and 480-nm images for each eye. No significant interocular differences in mean optic disc diameter were found. The average optic disc diameter across all eyes was 110 pixels. Thus, 100 pixels on the log E maps correspond to 1.6 mm at the retina or a 5.4° visual angle. 

The difference in log exposure, log (E(x,y)₄₈₀/E(x,y)₅₄₀) in the equation was calculated from the 480- and 540-nm log E maps (Fig. 2) for each pixel in a square area (8° × 8°) centered on the foveola. Profiles along a horizontal and a vertical line through the foveola were derived from this difference map. For the horizontal profile, the reference term, log (E(ref)₄₈₀/E(ref)₅₄₀), in the equation was derived from a linear interpolation between the values 4° nasal and 4° temporal to the foveola to account for nonuniformity in background reflectance caused by nonhomogeneous illumination of the retina and gradual changes in fundus pigmentation with eccentricity. These two terms were then substituted in the equation to yield a horizontal MP profile. The extinction coefficients K ₄₈₀ and K ₅₄₀ were derived from the MP spectrum of Bone et al. ²⁹ after accounting for the bandwidth of each filter. ¹⁹ The scaling term 1/(K ₄₈₀ − K ₅₄₀) was 1.39. A similar procedure was used to derive the vertical MP density profile by using superior and inferior reference values. On preliminary inspection, the vertical and horizontal profiles looked bell-shaped without sharp central peaks. Gaussian distributions (Fig. 3) were then fitted to the horizontal and vertical MP profiles. Areas with obvious specular reflections were ignored. We then calculated the peak density and the horizontal and vertical FWHMs. 

The measured MP density at any point was relative to that at the 4° eccentric reference site. MP density at 4° is 5% to 20% of the peak density. ^{12 37} Uneven illumination of the retina by the light source and reflections from blood vessels and the inner limiting membrane generally occur at eccentricities greater than 4°. Furthermore, the children’s photographs included the 4° eccentric site, even when the macula was close to the border of the field. 

Statistical Analysis

Statistical analyses were calculated by computer (SPSS, ver. 10.0; SPSS Inc.; Chicago, IL). Normality of distributions was evaluated with the Kolmogorov-Smirnov and Shapiro-Wilk tests. Comparison of distribution differences for large samples was determined with Student’s t-test and, for small samples, with the nonparametric χ² test. Within subject parameter consistency was determined by intraclass correlation coefficients. Results with P < 0.05 were considered significant. 

Results

Representative photographs (Fig. 1) from a 9-year-old (Table 1 , subject 7) show characteristics common to fundus images in all subjects. The center of the macula was darker in the 480- than the 540-nm image, demonstrating the absorption effect of the macular pigment. Images of the fundi obtained with 540-nm light were sharper than those obtained with 480-nm light (Fig. 1) , because of chromatic aberrations and the higher effect of light scatter at lower fundus reflectance. ³⁸ The density difference maps derived from the photographs in Figure 1 are shown in Figure 2 . The central bright area represents the high density of macular pigment. Shadow-like contours eccentric to the bright area represent blood vessels or other nonhomogeneous structures and arise from less than perfect alignment, or a difference in the sharpness of the 540- and 480-nm images, or both. The peak MP density and FWHMs, derived from the Gaussian fits (Fig. 3) for all 23 subjects are given in Table 1 . Across all eyes, peak MP density was 0.13 ± 0.04 DU, on average, and the width of the MP distribution (FWHM) was 2.4° ± 0.5°. 

For the 10 subjects who had both right and left eyes assessed, the intraclass correlations for the peak MP density (r = 0.06; P = 0.47) and horizontal FWHM (r = 0.1; P = 0.44) were not significant, but were significant for the vertical FWHM (r = 0.73; P = 0.02) and optic disc diameter (r = 0.65; P = 0.01). Although some have reported high correlation of MP parameters between right and left eyes, ³⁹ Bone et al. ³⁷ who included data from young children, found a considerable difference of 29% between the right and left eyes. For all eyes studied, no significant difference was found between the horizontal and vertical FWHMs. Furthermore, the horizontal and vertical FWHMs correlated significantly (right eye: r = 0.71, P = 0.001; left eye: r = 0.82, P = 0.001). Circular symmetry was demonstrated by a mean ratio between horizontal and vertical FWHMs close to unity (0.99 ± 0.06). 

The SD of peak MP density of the pooled data for right and left eyes was approximately 30% of the mean value, and the distribution of these densities was slightly skewed to lower values. Thus, the lower limit of normal MP density is at approximately the noise level. The standard deviations of the horizontal and vertical FWHMs were approximately 18% of the mean value skewed to higher values. Thus, the lower limit of normal for the FWHMs was more clearly defined than peak MP density. Neither peak MP density nor FWHMs varied significantly with age. Peak MP density did not correlate with FWHM. Repeated measurements of MP density in six eyes did not demonstrate significant differences (χ²; P = 0.67). The first and second measures correlated significantly (r = 0.77; P = 0.05). 

Discussion

Peak MP density in normal, young subjects (median age, 10.5 years), measured with monochromatic fundus photography, was 0.13 ± 0.04 DU, on average, with a minimum value of 0.07 DU and a maximum value of 0.18 DU. According to standard normal distribution, the variance of the data implied a range of 0.01 to 0.25 DU (P < 0.01). Peak MP densities found in the present study are slightly smaller, on average, than mean densities (0.2–0.3 DU) reported in studies using reflectometry. ^{17 18 19 21 31} Because of preretinal or intraretinal scatter and reflections, ^{15 19} MP densities are always smaller than MP densities obtained using psychophysical methods (mean density, 0.3–0.5 DU) ^{12 13 14 15 19} or autofluorescence spectrometry (0.5 DU). ¹⁹  

With respect to the variance of peak MP densities, those in the present study are similar to those found in other studies. For instance, the coefficient of variation (CV = 100 [SD/mean]) for peak MP densities in the present study is 27% for the right eye and 31.5% for the left eye (Table 1) . These coefficients are at the lower end of the range of CVs (8%–61%) obtained in other studies of MP density which used reflectometry, psychophysical, autofluorescence, or biochemical measurements. ^{12 13 19 40}  

It is unlikely that our MP densities are lower than those in other studies because of the young age of the subjects. No age effect was found among our subjects, aged 6 to 20 years, nor among other subjects within this age range. ^{14 19} Infants have lower MP densities. ³⁷ Thus, maturation of MP may occur before age 6 years. 

Factors other than age may have contributed to underestimation of MP density. First and most important, preretinal scatter—particularly, scatter in the crystalline lens—causes decreases in the contrast of all retinal features and in the estimated MP density. The illumination beam density in the lens increases with the area of the illuminated retinal field. The diameter of the illuminated area in our study was 30° compared with a maximum diameter of 3° in other reflectometry studies, including those in which confocal scanning laser ophthalmoscope devices were used. ¹⁹ Reflections at the limiting membrane, which are most intense in young individuals, ⁴¹ increase the variability of MP measurements. For the blue image, they decrease or increase the estimate of MP density at the foveola, or reference area (±4°). The opposite effects can be expected for the green image. Second, unbleached cone and rod photopigments (see the Methods section) could affect the MP density estimate. We evaluated this source of error and determined ¹⁹ that unbleached rhodopsin (<25%) would result in an underestimation of, at most, 0.005 DU, and unbleached cone photopigment (<7%) would result in an underestimation of, at most, 0.04 DU at 460 nm. Finally, we used a reference site 4° eccentric rather than 7° which was used in most other studies. This would result in an underestimation of the peak density by 5%, ¹² or, according to measurements on anatomic specimens, ³⁷ by as much as 20% to 25%. 

The absence of a significant interocular correlation between peak densities and FWHMs in the 10 subjects with measurements in both eyes is not consistent with most other studies. ^{12 19 37} Discrepancies may have been caused by the fact that, in the present study, only single values were used for left–right eye comparison rather than the average value of several individual measurements. Furthermore, specular reflections, particularly those extending over large areas, could not always be completely ignored in the Gaussian fits of the profile, and this undoubtedly contributed to errors. More elaborate procedures, such as the use of polarized light, could be developed to eliminate the effect of these specular reflections. 

Topographic distribution of the macular pigment in our subjects did not deviate substantially from circular symmetry. The ratio between horizontal and vertical FWHMs across all subjects was 0.00 ± 0.06; the FWHMs varied between 1.5° and 3.5° (450–1050 μm). Note that the FWHMs, derived from the approximated Gaussian profile, is not expected to vary appreciably when alternative models, such as an exponential fit, are used. ¹² Although the FWHMs of the present study demonstrate less variance, the mean FWHMs were slightly greater (2.4°) than that (2°) reported by Hammond et al. ¹² This discrepancy may be explained in part by our estimate of the optic disc diameter, which was as much as 10% smaller than the disc diameter after death. ³⁵ Perhaps because we used the latter as a reference, the mean FWHMs obtained in the present study, expressed in retinal distance (∼720 μm), is more consistent with those reported by Snodderly et al. ^{42 43}  

In summary, the photographic technique is feasible and reliable in pediatric subjects. A few minutes of cooperation are needed to obtain the photographic images for objective assessment of peak and topography of MP density. In view of the normative data reported herein, for pediatric patients the following would represent significant abnormalities: (1) Peak MP density so low that it is not measurable; (2) FWHM less than 1.5° (450 μm); and (3) horizontal and vertical FWHMs differing by more than approximately 25%. We anticipate increased sensitivity for detecting macular abnormalities if, in addition to peak MP density, the extent and circular symmetry of the MP density are also evaluated. With the advent of computerized fundus cameras and digital fundus photography, studies of the topography of MP density are expected to advance our knowledge of pediatric macular disorders. ^{3 4 5 6 7 8 9}  

 Supported in part by National Eye Institute Grant EY 10597.

 Submitted for publication July 13, 2001; revised January 25, 2002; accepted January 29, 2002.

 Commercial relationships policy: N.

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Fundus photographs of subject 7 (Table 1) . Top: Images of the left (OS) and the right (OD) fundi obtained with monochromatic green (540 nm) light. Bottom: images obtained with monochromatic blue (480 nm) light. The circular diameter is approximately 30°.

Figure 2.

 

Density difference maps derived from Figure 1 . The maximum difference, representing the macular pigment density, is visualized as a bright area near the center.

Figure 3.

 

Horizontal and vertical cross sections for the density difference maps in Figure 2 . The MP densities are double pass without spectral correction. The smooth curves are the fitted Gaussian functions.

Table 1.

 

Parameters of MP Density Distribution

Subject	Age	Peak MP Density (DU)			FWHM
								Horizontal (Degrees)			Vertical (Degrees)
				OD	OS	OD	OS	OD	OS
1	6	0.130		1.94		2.27
2	7	0.118	0.097	2.92	2.59	2.92	2.92
3	7		0.090		2.81		2.81
4	7		0.150		1.73		1.62
5	8		0.153		3.46		3.24
6	8	0.069	0.174	2.05	2.05	1.84	1.84
7	9	0.175	0.172	2.16	2.27	2.92	2.48
8	9	0.125		2.27		2.27
9	9	0.090	0.176	2.16	2.59	2.16	2.48
10	10	0.151		2.05		2.16
11	11	0.174	0.124	2.70	2.70	2.81	2.81
12	11	0.164		2.59		2.16
13	12	0.157	0.139	2.70	2.48	2.59	2.16
14	13		0.139		1.84		2.27
15	13	0.123		3.24		2.92
16	13	0.104	0.104	2.70	2.38	3.35	2.70
17	14	0.167	0.104	1.94	2.92	2.27	2.38
18	15	0.066		2.59		2.59
19	16	0.125		1.84		1.84
20	16		0.128		1.94		2.16
21	17	0.146	0.174	2.38	1.94	2.16	1.62
22	18		0.074		2.70		2.48
23	20	0.179	0.174	2.59	1.62	2.81	1.94
Mean	11.7	0.133	0.136	2.40	2.38	2.48	2.37
SD	4.3	0.036	0.034	0.39	0.50	0.42	0.47

The authors thank Pablo Artal (Laboratory de Optica, Universidad de Murcia, Murcia, Spain) for help with the calculation of the image quality of the 480- and 540-nm images.