Nonmydriatic fluorescence-based quantitative imaging of human macular pigment distributions

Abstract

We have developed a CCD-camera-based nonmydriatic instrument that detects fluorescence from retinal lipofuscin chromophores ("autofluorescence") as a means to indirectly quantify and spatially image the distribution of macular pigment (MP). The lipofuscin fluorescence intensity is reduced at all retinal locations containing MP, since MP has a competing absorption in the blue-green wavelength region. Projecting a large diameter, 488 nm excitation spot onto the retina, centered on the fovea, but extending into the macular periphery, and comparing lipofuscin fluorescence intensities outside and inside the foveal area, it is possible to spatially map out the distribution of MP. Spectrally selective detection of the lipofuscin fluorescence reveals an important wavelength dependence of the obtainable image contrast and deduced MP optical density levels, showing that it is important to block out interfering fluorescence contributions in the detection setup originating from ocular media such as the lens. Measuring 70 healthy human volunteer subjects with no ocular pathologies, we find widely varying spatial extent of MP, distinctly differing distribution patterns of MP, and strongly differing absolute MP levels among individuals. Our population study suggests that MP imaging based on lipofuscin fluorescence is useful as a relatively simple, objective, and quantitative noninvasive optical technique suitable to rapidly screen MP levels and distributions in healthy humans with undilated pupils.

Figures

Fig. 1

(a) Absorption spectrum of an excised, flat-mounted, human retina in the blue–green wavelength region, showing typical absorption characteristics of carotenoid macular pigment (MP) (solid curve at left). The retinal pigment epithelium (RPE) of the retina was removed for this measurement, and the spectrum was measured through a 1 mm aperture. The tissue absorption is remarkably similar in spectral shape to the absorption of a lutein/zeaxanthin solution, including the appearance of vibronic absorption substructure. The solid curve at right shows the fluorescence spectrum of a solution of lutein, obtained under excitation at 488 nm. (b) Molecular structure of lutein and zeaxanthin. (c) Energy level diagram of long-chain conjugated carotenoids such as lutein or zeaxanthin, with optical and nonradiative transitions indicated as arrows.

Fig. 2

(a) Absorption and emission spectra of a methanolic solution of A2E, the main fluorophore of lipofuscin, shown as solid curves at the left and right sides of the panel, respectively. The absorption transition between the lowest energy levels occurs in a broadband centered at 440 nm. Optical excitation leads to a strong emission centered at 660 nm. The overlapping absorption of MP is indicated as a dotted curve. The gray shaded area (λ >700 nm) indicates the lipofuscin fluorescence range used for indirect detection and imaging of MP distributions. (b) Molecular structures of (left) A2E and (right) iso-A2E. (c) Energy level diagram with optical transitions shown as arrows. Compared with carotenoids, the fluorescence transition is stronger in the A2E molecules due to their shorter conjugation length, which is interrupted by the central pyridinium ring.

Fig. 3

(a) Schematic representation of retinal layers participating in light absorption, transmission, and scattering of excitation and emission light in the macular region. ILM, inner limiting membrane; NFL, nerve fiber layer; HPN, Henle fiber, plexiform, and nuclear layers; PhR, photoreceptor layer; RPE, retinal pigment epithelium. In Raman scattering, the scattering response originates from the MP, which is located anteriorly to the photo-receptor layer. The influence of deeper fundus layers such as the RPE is avoided. In lipofuscin spectroscopy, light emission is generated from lipofuscin in the RPE layer to generate an intrinsic “light source” for single-path absorption measurements of anteriorly located MP layers. Using an excitation beam diameter exceeding the spatial extent of MP, the MP distribution can be quantified and spatially imaged (see the text). (b) Schematics of anterior optical media and retinal layers traversed by excitation laser light, fluorescence from anterior optical media, and fluorescence from lipofuscin. Iexc excitation light; IL, lipofuscin fluorescence; OM, ocular media; MP, macular pigment; PR, photoreceptors; RPE, retinal pigment epithelium; IOM, fluorescence from ocular media.

Fig. 4

Schematics of experimental setup used for imaging MP distributions in living human subjects. L1, L2, L3, lenses; F1, narrow bandwidth laser line filter; BS1, dichroic beam splitter; BS2, uncoated quartz beam splitter; AP, aperture; F2, F3, emission filters; CCD, charge-coupled detector array; PC, personal computer. A red aiming laser serves as a fixation target; depending on its direction, images can be recorded for either the macular region or the peripheral retina.

Fig. 5

Removal of laser speckle effects in fluorescence imaging, illustrated for [(a) and (b)] gray-scaled images of millimeter graph paper and [(c) and (d)] MP distributions. Granularity of images is removed in (b) and (d) when scrambling the excitation light via fiber mode mixing. This makes it possible to resolve fine details in the retinal images such as small retinal blood vessels at the periphery of the macula [(d)]. Calibrating the image dimensions with the known pixel array dimensions (1 mm equals 120 pixels), we see that retinal blood vessels as thin as ~20 μm can be clearly resolved in the used setup.

Fig. 6

Determination of spatial correction factor for image processing due to spatially varying intensities in laser excitation and lipofuscin fluorescence disks. (a) Lipofuscin fluorescence intensity in the peripheral retina, plotted versus distance along horizontal meridional line running through the center of a ~3.5-mm-diameter image disk (solid curve). The intensity profile has a good match to a Gaussian fit (dashed curve), as expected for the Gaussian intensity profile of the exciting laser disk. At the edge of the spot the intensity has dropped by ~20% with respect to the center of the profile. This value is used as a correction factor. (b) and (c) Schematics of central and peripheral retinal images recorded sequentially for measured eyes. In image (b), the excitation disk is centered on the fovea F, and in image (c), it is centered on a spot S in the peripheral retina that is 7° eccentric. Spot S is visible in both images. Since the spot has a constant lipofuscin concentration, the fluorescence intensities originating from it in images (b) and (c) have to be due to the excitation intensity profile, and the intensity difference can be used as a spatial correction factor in the determination of MP levels from image (b) or (c).

Fig. 7

(a) White-light fundus camera image of a healthy subject (a), and lipofuscin fluorescence images of the fundus obtained at near IR wavelengths with (b) 488 nm and (c) 532 nm laser excitation. Note the pronounced attenuation of fluorescence in the center of image (b) due to MP absorption and the nonattenuated fluorescence in image (c). Pixel intensity maps of images (b) and (c) are used for an indirect derivation of MP concentration levels existing in the macular region (see the text).

Fig. 8

Schematics showing processing of CCD pixel intensity regions to derive optical density values of MP absorption at any desired location in the retina. Individual pixels are grouped into disks with a diameter of 20 pixels each, and the intensities in these disks are averaged. One central disk is located at the center of the macula, the faveola, with a resulting intensity Imin(ave). Twelve additional disks are chosen on a circle with 7° eccentricity to the foveola, with equidistant spacing, to calculate an average fluorescence intensity Imax(ave) in the periphery. The maximum MP image contrast, derived from these two averaged intensities, is proportional to the optical density of MP in the center of the macula, according to Eq. (10). Disks between the center and the peripheral circle (not shown) can be chosen to calculate the image contrast and MP at any eccentricity toward the peripheral retina.

Fig. 9

Measurement of lutein concentrations in a tissue phantom consisting of a dried lutein spot located on the side wall of a thin cuvette filled with an optically thin A2E solution (O.D. ~0.35). Optical densities of lutein concentrations were measured for several dozen positions within the lutein spot, using simultaneous lipofuscin fluorescence attenuation and transmission measurements of the individual spots. The dried lutein spot increased in thickness toward its periphery, thus providing a test sample with increasing strongly varying absorptions. Setup shown (a) schematically in side view and (b) from top. (c) Lutein optical densities measured with both techniques, showing excellent correlation (correlation coefficient R=0.96). Transmission-based absorption levels had an accuracy of 2%; indirect, lipofuscin-based absorption levels had an accuracy of 4%.

Fig. 10

Wavelength dependence of MP optical density obtained when successively limiting the detected fluorescence range to longer wavelengths, using long-wavelength pass filters with suitable cut-on wavelengths λc. The effect was measured for three subjects. Optimum image contrast and resulting maximum MP optical density are obtained when limiting the detection of lipofuscin fluorescence to wavelengths λ≥700 nm. Note the strong reduction of MP optical density, caused by decreased image contrast when including green–yellow wavelengths. The effect is caused by artifactual emission signals in that wavelength region originating from the human lens.

Fig. 11

Lipofuscin fluorescence images of four volunteer subjects [(a)–(d)], obtained under 488 nm excitation, shown in gray scale. Fluorescence intensities are lowest in central dark image regions due to absorption of excitation light by MP. Note the pronounced variation of MP distributions regarding strength, symmetry, and spatial extent among individuals.

Fig. 12

Lipofuscin fluorescence intensity profiles along nasal–temporal meridians (solid curves) and inferior–superior meridians (dashed curves), all running through the center of the macula (see the inset). Plots are derived from pixel intensity maps of Fig. 8 for each of four subjects [(a)–(d)]. A transmission value in the foveola can be calculated from the averaged pixel intensities in the peripheral macula and the foveola.

Fig. 13

MP optical density plot profiles along horizontal axis, derived from Fig. 11 for each subject.

Fig. 14

Pseudocolor-scaled, three-dimensional MP distribution derived from two-dimensional lipofuscin fluorescence pixel intensity maps of Fig. 8. Note the significant intersubject variations in MP levels, symmetries, and spatial extent.

Fig. 15

Effect of blood vessels on transmission line plots derived from gray-scale lipofuscin fluorescence intensity maps. Two lipofuscin fluorescence images are compared: (a) a standard gray-scale image and (b) the same image convoluted with a mask filter, revealing finer details. For both images, nasal–temporal transmission plots are calculated and shown in (c) and (d), respectively, for rectangular pixel areas of 5 pixels height and 150 pixels width (dashed curves) and 15 pixels height and 150 pixels width (solid curves). The appearance of small blood vessels on the order of ~50 μm diameter (~2 pixels) is clearly seen in the wings of the MP profile. The spatial asymmetry of the profile, however, is not influenced by the blood vessels. Note the small dip in absorption of MP in the faveola for this subject.

Fig. 16

(a) Categories of MP distributions observed in clinical measurements of 70 subjects. Two examples are shown for each category. Category A has very low MP levels (O.D. smaller than 0.05), B features enhanced central MP levels and lower eccentric levels, C has only a sharp, central MP distribution, D has a central MP concentration surrounded by lower amounts arranged as a ring or shoulder, and E has both central and parafoveal MP levels. (b) Distribution of 122 measured eyes among categories A–E. A large fraction of subjects, 28%, has a sharp, central MP distribution.

Fig. 17

MP levels measured in 90 eyes of a clinical population of 70 volunteer subjects, displayed as a function of age. Each data point corresponds to the maximum MP level determined from lipofuscin fluorescence images. Note the significant variation of MP levels among individuals at any age and the average decline of levels with age. Solid circles represent subjects measured after cataract surgery (lens implant).

Fig. 18

Correlation between MP levels determined by lipofuscin-fluorescence-based imaging and resonance Raman spectroscopy. 72 eyes of 48 subjects were measured for these experiments. Fluorescence-based MP optical density levels are maximum levels at the foveola, calculated from lipofuscin fluorescence pixel intensity maps integrated and averaged for a circular area, centered at the fovea, having 20 pixel diameter (150 μm). Raman-based MP levels are derived from Raman scattered light intensities at the carbon double stretch frequency of 1525 cm−1. Since the latter are obtained with a 1-mm-diameter, 488 nm laser, excitation spot centered on the macula, Raman-based MP levels are averaged over a 1-mm-diameter area. A remarkably high correlation (correlation coefficient r=0.73) is obtained between both methodologies. Raman-based levels are not corrected for media opacities, while fluorescence-based levels are not influenced by media opacities. The high correlation therefore demonstrates indirectly that media opacities are not very significant in Raman measurements.