Macular pigment: quantitative analysis on autofluorescence images

Abstract

Background

Macular pigment (MP) reduces oxidative damage in the central retina and can be quantified by flicker-photometric analysis (HFP) of MP optical density. These analyses demonstrate a very good correlation with central absorption by MP on autofluorescence (AF) images. With these techniques different types of MP-distribution have been described. In the present study a quantification analysis of MP in AF images was developed to verify these MP types and to compare MP distribution patterns between healthy individuals and those with age-related macular degeneration (AMD).

Methods

AF images (HRA) were analysed with respect to the area of central and paracentral absorption in 400 eyes with a computerised analysis program of MP optical density. The patients were between 41 and 90 years old (mean 67.2 years); 168 were male and 232 female, and 253 had early AMD and 147 showed no AMD characteristics. The central MP concentrations (peak) were measured, the amount of MP values within the first 8-pixel radius (“C”), the total amount of MP within a 120-pixel radius (“T”) were calculated as the volume of the MP values over the regarded radius and the C/T ratio was registered.

Results

Four types of MP distribution (type 1, intense central and paracentral MP; type 2, less intense central and paracentral MP; type 3, only central MP; type 4, only paracentral MP) were identified. The differences in MP distribution were confirmed and clearly characterised by quantitative analyses of peak, total MP (“T”), central MP (“C”) and C/T ratio: mean peak in type 1, 0.65; type 2, 0.42; type 3, 0.42; type 4, 0.29; mean total amount of MP in 120-pixel radius (“T”) in type 1, 5829.0; type 2, 4412.5; type 3, 2709; type 4, 4302.8. MP types with lower levels of MP were significantly more often observed in the AMD group (AMD: type 1, 120=47.4%; types 2–4, 133=52.6%; healthy eyes: type 1, 112=76.2%; types 2–4, 35=23.8%) (P<0.0001)

Conclusions

Analysis of MP on AF images is a quantitative method for investigation of MP. With this method a wide variation in concentration and distribution of MP could be seen in the population. Four different types of MP distribution could be characterised and quantitatively distinguished. Reduced levels of MP seem to be associated with a higher risk of development of AMD as they were significantly more often observed in the AMD group. This strategy of quantitative MP analysis on AF images is easily practicable and may be used in further studies to investigate the role of MP as a potential risk factor for AMD.

Appendix: Analysis

Let F _F (Λ,λ) and F _P (Λ,λ) be the AF (Λ is the excitation wavelength, λ the emission wavelength) measured at the fovea and the perifovea, let D _F and D _P be the optical density of MP at both sites, and F* _F (Λ,λ) and F* _P (Λ,λ) be the fluorescence of all layers located posteriorly to the MP. We assume that all detected fluorescence has been affected by MP absorption and modify Delori’s formula as follows (as due to the broad band-pass barrier filter we measure all wavelengths from 500 to 720 nm that reach the detector, not just a single wavelength as used in recent studies [14] Furthermore, we consider the spectral responsivity S(λ) of the detector. Thus we can express the foveal and perifoveal fluorescence as:

$$ \begin{array}{*{20}l} {{F_{F} (\Lambda ) = {\int_{500}^{720} {S(\lambda )*F^{*}_{F} (\Lambda ,\lambda )*10^{{ - D_{F} (\Lambda ) - D_{F} (\lambda )}} d\lambda = } }} \hfill} \\ {{10^{{ - D_{F} (\Lambda )}} *{\int_{500}^{720} {S(\lambda )*F^{*}_{F} (\Lambda ,\lambda )*10^{{ - D_{F} (\lambda )}} d\lambda } }} \hfill} \\ \end{array} $$

(1)

$$ \begin{array}{*{20}l} {{F_{P} (\Lambda ) = {\int_{500}^{720} {S(\lambda )*F^{*}_{P} (\Lambda ,\lambda )*10^{{ - D_{P} (\Lambda ) - D_{P} (\lambda )}} d\lambda = } }} \hfill} \\ {{10^{{ - D_{P} (\Lambda )}} *{\int_{500}^{720} {S(\lambda )*F^{*}_{P} (\Lambda ,\lambda )*10^{{ - D_{P} (\lambda )}} d\lambda } }} \hfill} \\ \end{array} $$

(2)

Now we determine the logarithm of the quotient of Eq. 2 and Eq. 1:

$$ Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)} = D_{F} (\Lambda ) - D_{P} (\Lambda ) + Log{\left( {\frac{{{\int_{500}^{720} {S(\lambda )*F^{*}_{P} (\Lambda ,\lambda )*10^{{ - D_{P} (\lambda )}} d\lambda } }}} {{{\int_{500}^{720} {S(\lambda )*F^{*}_{F} (\Lambda ,\lambda )*10^{{ - D_{F} (\lambda )}} d\lambda } }}}} \right)} $$

(3)

Applying Beer’s law we get:

$$ Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)} = D_{{AF}} (460)*K_{{MP}} (\Lambda ) + Log{\left( {\frac{{{\int_{500}^{720} {S(\lambda )*F^{*}_{P} (\Lambda ,\lambda )*10^{{ - D_{P} (\lambda )}} d\lambda } }}} {{{\int_{500}^{720} {S(\lambda )*F^{*}_{F} (\Lambda ,\lambda )*10^{{ - D_{F} (\lambda )}} d\lambda } }}}} \right)} $$

(4)

If we neglect the log term ($F^{*}_{P} (\Lambda ,\lambda ) \approx F^{*}_{F} (\Lambda ,\lambda )$ and D _P (λ)≈0 and D _F (λ)≈0 for most of the detected wavelengths) on the right side we get:

$$ Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)} = D_{{AF}} (460)*K_{{MP}} (\Lambda ) $$

(5)

or

$$ D_{{AF}} (460) = Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)}*\frac{1} {{K_{{MP}} (\Lambda )}} $$

(6)

If we consider the extinction coefficient for Λ=488 nm we get:

$$ D_{{AF}} (460) = Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)}*1.3 $$

(7)

Let C be the brightest spot on the image. Applying the rules of logarithm we can calculate:

$$ \begin{array}{*{20}l} {{D_{{AF}} (460) = Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)}*1.3 + 0*1.3 = } \hfill} \\ {{Log{\left( {\frac{{F_{P} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)}*1.3 + Log{\left( {\frac{{F_{C} (\Lambda )}} {{F_{C} (\Lambda )}}} \right)}*1.3 = } \hfill} \\ {{Log{\left( {\frac{{F_{C} (\Lambda )}} {{F_{F} (\Lambda )}}} \right)}*1.3 - Log{\left( {\frac{{F_{C} (\Lambda )}} {{F_{P} (\Lambda )}}} \right)}*1.3} \hfill} \\ \end{array} $$

(8)

This means we can calculate the optical density relatively to the brightest spot in each image. Then we have to subtract the optical density of the plateau from this value to get the optical density of the fovea relative to the plateau. First we determine the radial distribution of the optical density around the fovea. Since we want the location of the plateau to be variable, we normalise the optical densities to the brightest spot in the image and display it.

Now the plateau is marked. Using Eq. 8 we subtract the optical density of the plateau to get the values relative to the plateau. These values are evaluated, averaged, integrated and so on.