Fundus autofluorescence in age-related macular degeneration
Jose M Ruiz-Moreno, MD, PhD
Department of Ophthalmology, Vissum Alicante & CHUA. Spain.
Alicante Institute of Ophthalmology, VISSUM, Vitreo-Retina Unit. Alicante. Spain.
Javier A Montero, MD, PhD
Pio del Rio Hortega Hospital, University of Valladolid. Valladolid. Spain.
Alicante Institute of Ophthalmology, VISSUM, Vitreo-Retina Unit. Alicante. Spain.
Virginia Bautista Ruescas, MD
Pio del Rio Hortega Hospital, University of Valladolid. Valladolid. Spain.
The authors have neither economical nor commercial interest in the devices and procedures described.
The authors wish to thank TOPCON Spain SA for its help providing the Spaide AF filter Set and adapting it to the Topcon TRC-50DX camera.
All the images presented in this chapter were obtained using this system.
Fluorescence is the capability of absorbing light at a specific wavelength and releasing it at a longer, less energetic wavelength.
This phenomenon raises an especial interest when the released radiation is found within the spectrum of visible light permitting its visualization, recording and measurement.
Autofluorescence is the spontaneous fluorescence that some substances present naturally.
Fluorophore is the part of a molecule that makes it fluoresce.
The human eye contains autofluorescent substances in the retina, especially within the retinal pigmentary epithelium (RPE).
The main autofluorescent component of the RPE is lipofuscin (LF), containing at least ten difffferent fluorophores presenting discrete emission spectra within the green, golden-yellow, yellow-green, and orange-red emitting range(1).
When LF granules are stimulated with light within the blue range, a characteristic yellow fluorescence is emitted(2).
Basic considerations on fundus autofluorescence
The RPE plays an important role in the physiopathology of age-related macular degeneration (AMD)(3, 4).
The study of RPE can help to achieve a better understanding of AMD and to find new ways of early diagnosis as well as new prognostic and progression markers for this condition.
LF originates from the constant phagocytosis of the shed outer segment disks of the photoreceptors and is accumulated in the cytoplasm of RPE cells(5,6).
This accumulation is considered to be a hallmark of RPE aging(7).
Recent experimental studies have addressed the molecular mechanisms of the interaction of excessive LF with the normal cellular functions of RPE(8,9).
According to these studies, A2-E (N-retinylidene-N-retinylethanol-amine) has been identified as the main fluorophore of LF.
A2-E may play a toxic effect including phototoxic and detergent actions, as well as an inhibitory effect on lysosomal function(10,11).
It has been suggested that the photooxidation by products related to LF may trigger the complement cascade, thus contributing to the pathological chronic inflammation of the macular area(12).
New studies have been designed to improve our understanding of potential underlying molecular mechanisms.
The autofluorescence of LF, its distribution in post-mitotic human RPE cells and its accumulation with age have been extensively studied in post-mortem eyes with fluorescence microscopy(14,15).
In the past few years we have started to study the autofluorescence in vivo.
Ultraviolet light is frequently used to visualize LF by fluorescence microscopy ex vivo, since the absorption properties of the eye limit the transmission of ultraviolet light within the retina in human living eyes.
However, due to the wide range of excitation of LF (from 300 nm to 600 nm), visible light can be used to visualize its fluorescence in vivo.
The emitted spectrum ranges from 480 to 800 nm and peaks within the range of 600 to 640 nm(15).
Fundus autofluorescence (FAF) imaging of the human living eye is a relatively new imaging method that provides a topographic map of the distribution of LF in the RPE.
The detection of FAF is limited by its low intensity (approximately two orders of magnitude lower than the peak background fluorescence of an ordinary fluorescein angiography), and by the autofluorescence characteristics of the anatomic structures of the eye, including those of the optical media, especially of the lens(16).
Fundus spectrophotometry was developed by Delori et al. and was designed to determine the spectrum of excitation and fluorescence emission from small areas of the retina (2º diameter)(17,18).
The authors were able to determine the amount of autofluorescence and compare it with in vitro fluorescence microscopy.
They found out that the spectrum of in vivo excitation was slightly broader, and peaked at a longer wavelength than those of A2E and native LF.
The authors concluded that considering the spatial distribution, spectral characteristics and age relationship, LF is the main source of fluorescence in the FAF in vivo(18).
Presently, two systems are available to examine the autofluorescence of the human eye in vivo in the clinical practice: confocal scanning laser ophthalmoscope and fundus camera.
Scanning laser ophthalmoscope
Confocal scanning laser ophthalmoscope (cSLO) was originally developed by Webb et al. using a low-energy laser source to scan the retina in two directions: termed as x and y(19).
The confocal nature of the optics ensures that the reflectance and fluorescence correspond to the same focal plane. cSLO overcomes the limitations of the low-intensity signal of FAF and the lens interferences.
The defocused light is almost completely suppressed, thus reducing the autofluorescence from the optical media anterior to the retina, such as the lens or the cornea.
In order to reduce the background noise and to increase the contrast of the image, a series of FAF images are usually recorded(20,21).
Following the aligning of the images in order to correct the movement of the eye during the acquisition, the final image is calculated (usually from 4-32 frames) and the values of the pixels are normalised.
The FAF image can be obtained with low excitation energies within the limits of maximum retinal irradiance established by the American National Standard Institute and other international standards(22).
cSLO enables the acquisition of FAF images from wide areas of the retina (55º with one frame and even larger areas using the composite mode)(20,22).
Although limited by the optical properties of the human eye, SLO succeeds in imaging the posterior pole with a high contrast.
Currently there are three different cSLO systems for FAF imaging: the Heidelberg Retinal Angiograph (HRA) (based on the HRA classic, HRA 2 and the Spectralis HRA) (Heidelberg Engineering, Dossenheim, Germany); the Rodenstock cSLO (RcSLO; Rodenstock, Weco, Düsseldorf, Germany); and the Zeiss prototype SM 30 4024 (ZcSLO; Zeiss, Obercochen, Germany).
HRA is the only currently commercially available system with the cSLO system to capture FAF images.
HRA uses an excitation wavelength of 488 nm from an Argon laser or a solid-state laser. A barrier filter with a short-wavelenght cutoff at 500 nm is inserted just opposite the detector, blocking the laser light and letting the autofluorescent light through.
Recently, it has been made possible to acquire real time images, a technique known as real-time averaging.
The Rodenstock cSLO and the Zeiss prototype SM 30 4024 have also been used to acquire clinical FAF images.
Both systems use an excitation wavelength of 488 nm (the same as the HRA), and barrier filters at 515 nm and 521 nm, respectively(20,21,23).
Bellmann et al. have noticed marked contrast and brightness differences as well as in the grey range (an important marker of the image quality between the different systems of cSLO).
These limitations must be taken into consideration when comparing images from different cSLO systems(24).
The default software of the HRA system normalizes the pixel distribution of the final image in order to improve the distribution of the FAF intensity.
Even though this final step facilitates the evaluation of the localized topographic differences, it allows a relative estimation of the intensities of the FAF.
Thus, it should not be used for quantitative calculation and absolute comparison between different FAF images.
The normalization of the average images can be easily turned off, and brightness and contrast can be manually adjusted to permit an adequate visualization of the distribution of autofluorescence in areas with a very high or very low signal in order to improve the visualization of small details.
Fundus cameras are widely used in clinical routine for imaging the retina as fundus photographs, reflectance photographs and fluorescein angiography.
Fundus cameras use a single flash to capture images from large retinal areas.
When confocal optics are not available, the autofluorescent signal from all the ocular structures with fluorescent properties reaches the camera, and scattered light, anterior and posterior to the plane of interest can influence the detected signal(25-27).
The lens contributes significantly to the autofluorescent signal when similar wavelengths are used in the blue-range, as for the cSLO (λ = 488 nm), particularly in older patients with lens yellowing and nuclear opacities.
Flashlight intensities and detector gain have to be set at relatively high levels in order to obtain reasonable FAF images.
However, the signal-to-noise ratio decreases simultaneously which may result in reduced image quality.
To reduce interference from lens fluorophores which mainly emit in the range between 510 to 670 nm, Spaide modified the excitation filter (peak 580 nm, bandwidth 500-610 nm) and the barrier filter (peak 695 nm, bandwidth 675-715 nm).
A further modification was introduced in 2007 using a slightly different filter set (excitation bandwidth 535-580 nm, emission bandwidth 615-715 nm)(28), thus improving signal-to-noise ratio and image quality.
Furthermore, as this setup of the fundus camera uses different excitation and emission filters compared with the cSLO, it may even visualize other retinal fluorophores.
However, a systematic comparison of different pathologies with clinico pathological correlations between cSLO and fundus camera, particularly in patients with AMD, has not yet been performed.
Originally, a fundus camera that enabled imaging with a field of 13º was used.
Recently, Spaide has obtained images of the spatial distribution of FAF intensities over larger retinal areas up to 50º with his new modified fundus camera(25-28).
In the near future we may improve FAF imaging with the aid of scientists and investigators developing filters and some other innovations, and increased experience.
Furthermore, it is already possible to visualise different fluorophores from the retina with the configuration of the fundus camera using excitation and emission filters for the cSLO.
A systematic comparison of clinical images with different pathologies obtained by the cSLO and the fundus camera, (especially in AMD patients) has not been performed yet.
Autofluorescence imaging in the human eye in vivo
FAF images show the spatial distribution of the intensity of autofluorescence of each pixel in grey values (arbitrary values from 0 to 225); low intensities are commonly known as low pixel values (dark) and high intensities as high pixel values (light).
FAF imaging shows a consistent pattern of autofluorescence distribution in normal eyes(21).
Such common findings have been reported in children as young as four years old(29).
The macular FAF signal is reduced at the fovea because it is limited by the presence of lutein and zeaxanthin in the neurosensory retina.
The signal is higher in the parafoveal area and tends to increase as we move away from it, peaking at the most peripheral retinal areas.
It has been suggested that this FAF pattern is caused by the melanin deposition and density of LF granules at the different areas of the retina(18,30).
The optic nerve head typically appears dark mainly due to the absence of RPE.
The retinal vessels are associated with a markedly reduced FAF signal because of the blocked fluorescence (Fig.1).
The common ratios of grey intensity between the fovea and the perifoveal area have been established(31,32).
Considering these findings, any deviation from the normal pattern in a specific location can be easily identified; hence the qualitative description of the local changes in the FAF is widely used.
The changes in signal intensity are qualitatively described as decreased, normal, or increased as compared to the background signal of the same eye.
FAF imaging in AMD
When examined with autofluorescence, the fundus of patients with AMD may show a range of signal changes(20,33-37).
Assuming that RPE has an important role in the pathophysiology of AMD and that the major fluorophores in the retina are located within RPE cells, FAF imaging can show changes in the concentration and distribution of RPE LF and hence establish the condition of RPE in patients with AMD.
All this information can be obtained from a quick and minimally invasive exploration with FAF.
The decreased FAF intensity may also be associated with hyperpigmented areas due to the melanin absorption of light(35,39).
However, it should be considered that other fluorophores than LF can be found in RPE and become more prominent in AMD patients, and hyperpigmented areas may also cause an increase in the signal, which is supposed to result from the accumulation of melanolipofuscin.
Other changes in FAF which are not related to RPE defects may appear in AMD.
Fresh haemorrhages typically appear dark due to blocked fluorescence (Fig. 3).
Pigment epithelial and neurosensory detachment and areas with extracellular fluid accumulation associated with exudative lesions can be observed in FAF as increased or decreased signal intensity.
Fluid accumulation under pigment epithelium detachment, extracellular deposition of material under the RPE (drusen), and fluid originated from CNV can occur with increased, normal or decreased FAF intensity.
This phenomenon is a consequence of the presence of unknown autofluorescent molecules other than LF, in the same spectral range than LF.
FAF imaging alone may not distinguish between melanolipofuscin from RPE cells migrated into the neurosensory retina and LF within the normal RPE layer.
It is always necessary to compare the FAF findings with those from other techniques such as fundus photograph, reflectance image, fluorescein angiography or optical coherence tomography (OCT)(35, 39).
FAF in early AMD
Early AMD is characterised by the appearance of localized RPE hypo or hyper pigmentation and drusen.
Drusen are formed by the accumulation of extracellular deposits in the inner aspects of Bruch’s membrane(3).
Depending on their size and morphology, they can be classified as hard or soft drusen.
The molecular composition of drusen is quite complex and has not been completely elucidated (Fig. 5).
FAF changes in early AMD have already been reported by several authors (9,21,25,27,33-36,41); all of them concluding that the changes in ophthalmoscopy and fluorescein angiography are not necessarily related with FAF, suggesting that FAF may provide new information regarding the stages and activity of the disease.
Differentiation between RPE LF and sub-RPE deposits with FAF images in vivo can be a hard work.
An analysis of the variability of FAF in patients with early AMD was recently reported by an international workshop on FAF phenotype in early AMD.
Among their conclusions, a new classification system with eight different FAF patterns was given(39).
Normal pattern characterized by a homogeneous background autofluorescence with a gradual fluorescence decrease in the inner macula towards the foveola (blocked fluorescence caused by yellow macular pigments).
FAF may be normal even in the presence of soft or hard drusen. Minimal change pattern characterized by a limited and irregular decrease or increase of background AF, not associated to any obvious or important topographic pattern.
85 Fundus autofluorescence in age-related macular degeneration Focal increased pattern is defined by the presence of at least one well defined spot (<200 micron diameter) of markedly increased FAF much brighter than the surrounding background fluorescence.
These areas may or may not correspond to large, soft drusen and to areas of hyperpigmentation.
Patchy pattern characterized by the presence of at least one large area with well defined borders (>200 micron diameter) of markedly increased FAF.
Again, these areas may or may not correspond to large, soft drusen and areas of hyperpigmentation.
Linear pattern defined by the presence of at least one linear area of markedly increased FAF with well defined borders.
Linear structures of increased FAF usually correspond to hyperpigmented lines.
Lacelike pattern This pattern shows multiple branching lines of increased FAF forming a lacelike pattern.
The borders may be hard to define, and FAF may gradually decrease from the centre of the linear area towards the surrounding background.
This pattern may correspond to hyperpigmentation or to non visible abnormalities.
Reticular pattern is defined by the presence of multiple small, ill defined areas (<200 micron diameter) of decreased FAF.
This pattern has been found to occur not only in the macular area, and may be associated with multiple small soft drusen, hard drusen or areas with pigment changes or non visible abnormalities.
Speckled pattern is characterized by the simultaneous presence of different types of abnormalities in a large area.
The changes reach beyond the macular area and may cover the entire posterior fundus.
These abnormalities include multiple small areas of irregular increased or decreased FAF corresponding to hyper and hypopigmented areas and multiple subconfluent and confluent drusen.
The speckled pattern has been reported to be the most frequent (26%) followed by the patchy pattern (23%).
The most infrequent patterns are the normal pattern (2%) and the lacelike pattern (2%).
The study confirmed that visible drusen on fundus photography are not always correlated with noticeable FAF changes and that areas of increased FAF may or may not correspond to areas of hyperpigmentation or soft or hard drusen.
Several authors have also mentioned the different FAF patterns in eyes with drusen.
Delori et al. described a pattern consisting of decreased FAF in the centre of the drusen surrounded in most of the cases by a ring of increased FAF(27).
They also observed that the decreased drusen signal was not as intense as in the areas with RPE atrophy.
The authors hypothesized that it might be caused by a displacement of the cytoplasm and LF granules in RPE cells instead of an actual RPE atrophy(35).
Von Rückmann et al. further reported that crystalline drusen are characterised by a decrease in FAF signal, signalling the onset of atrophy.
Lois et al. confirmed that areas of confluent drusen are usually associated with focal, mildly increased FAF and that only large subfoveal soft drusen (drusenoid RPE detachments) topographically correspond with focal changes of FAF(33) (Fig. 6).
Smith et al. recently reported their results after using image analysis software to study drusen and pigmented areas on fundus photographs from AMD patients(41).
The authors initially used image analysis algorithms, including automated background levelling and thresholding.
Areas of focally increased FAF intensities were compared to the normal background signal.
By overlapping fundus photographs and FAF, the topographic correlation of drusen and pigmented areas with focally increased FAF signals was established.
Smith and co-workers reported that eyes with isolated drusen or pigment abnormalities were better correlated with FAF abnormalities than eyes with geographic atrophy(41).
Regarding areas with changes in RPE, hypopigmented areas are usually associated with a corresponding decreased FAF signal, suggesting an absence or degeneration of RPE cells, with reduced content of LF granules. (Fig. 7 and 8).
Advanced AMD is characterized by geographic atrophy (GA), choroidal neovascularization (CNV), pigment epithelial detachment (PED), RPE tears and disciform scars.
Geographic atrophy is thought to be the natural end stage of the atrophic AMD process when CNV does not appear.
GA occurs in areas where the RPE is dead and the outer neurosensory retina and choriocapillaris disappeared(42, 43).
Due to the loss of RPE and LF, the atrophic area appears dark in FAF imaging(35).
High contrast between the atrophic and the non atrophic retina defines the area of GA more precisely than colour fundus photographs, permitting a clearer and more specific study of GA, as well as its natural development and evolution(38,44) (Fig. 10 and 11).
The GA patches usually become larger and coalesce as AMD progresses(45,46).
An excessive accumulation of LF, and therefore an increased FAF in the junction are highly suggestive of the appearance or progression of pre-existing GA (Fig. 12).
Preliminary observations suggest that different phenotypes may appear associated with junction FAF changes(47).
-Focal increased autofluorescence is defined by single or multiple spots of focal markedly increased FAF localized at the border of the atrophic patch.
-Band pattern of increased autofluorescence is characterized by a continuous stippled band of increased FAF surrounding the entire atrophic area.
-Patchy increased autofluorescence are large patches of increased FAF outside the GA area.
FAF tends to be less intense than that in the focal pattern described above. -Diffuse increase autofluorescence is the most frequent pattern of incresed FAF in eyes with GA.
FAF changes are not limited to the border of the atrophic area and may show inter individual differences that have been further classified into four subtypes.
-Reticular pattern, characterised by several lines of increased FAF usually following a radial pattern.
-Branching pattern shows a diffusely increased FAF with a fine branching pattern of increased FAF.
-Fine granular pattern, is defined by a large area of increased FAF with a granular appearance surrounding the GA area and a clear border between the granular increased FAF and the surrounding normal background FAF.
-Fine granular with peripheral punctate spots pattern is characterised by diffuse FAF changes surrounding the atrophic area with elongated small lesions and increased FAF.
Refined phenotypes help to identify the prognosis and seem to be a prerequisite to determine specific genetic factors in a complex, multifactorial disease such as AMD.
A recent analysis of the follow-up of junction FAF patterns in GA and the rate of progression of atrophic lesions revealed that variation in GA growth rates are dependent on the specific phenotype of FAF at baseline(49).
Atrophy enlargement was slowest in eyes with normal FAF pattern (median, 0.38 mm2/year), followed by focal FAF pattern (median, 0.81 mm2/year), diffuse FAF pattern (median, 1.77 mm2/year), and banded FAF pattern (median, 1.81 mm2/year).
The rate of progression of GA in eyes with patchy FAF pattern were not included in this analysis because of their low frequency, insufficient for statistical analysis.
The rate of progression of “banded” and “diffuse” FAF patterns were significantly higher compared to eyes without FAF abnormalities and “focal” FAF pattern.
Another interesting finding of this study was the identification of eyes with extremely rapid progression of the atrophy, showing distinct FAF features of atrophy that had not been previously reported.
The authors introduced the term “diffuse trickling” for a pattern associated with a significantly faster enlargement of atrophy.
Areas with increased FAF and consequently higher concentrations of RPE LF precede the development of new areas of GA or the enlargement of the pre existing atrophic areas.
The phenotypic features of FAF abnormalities may play a stronger influence on the progression of atrophy than any other previously reported risk factors such as smoking, arterial hypertension or diabetes.
The different rates of enlargement of atrophy may be related to heterogeneity at a cellular and molecular level in the disease.
The high degree of symmetry in GA suggests that genetic determinants may be involved, rather than nonspecific aging changes.
Choroidal neovascularization is considered to cause almost 90% of the cases of severe visual loss related to AMD(50).
CNV is usually studied by fluorescein angiography and OCT to assess the extent, location and nature of the lesion(51).
Fluorescein angiography shows changes in retinal vascularization, but does not reveal how deeply RPE is affected.
FAF imaging shows RPE damage, with the advantage that is a non invasive test, less time consuming than angiography.
Areas of abnormal FAF extend beyond the edge of the angiographically defined lesion.
As in other exudative retinal diseases such as central serous chorioretinopathy, areas of increased FAF next to CNV are frequently found inferior to the leaking areas in fluorescein angiography.
The hypothesis was that they might represent areas with subretinal fluid and that their location was influenced by gravity. Other fluids may typically decrease FAF, as occurs with haemorrhages and exudates. Decreased FAF is caused by blocked fluorescence.
It is usually necessary to compare the results of FAF with colour photographs.
Recent research has examined early CNV in FAF(36, 52, 53), reporting that early CNV lesions tend to show normal FAF in areas that were hyperfluorescent in fluorescein angiography, whereas eyes with a history of one month or more since CNV was diagnosed, showed decreased FAF in areas of previous fluorescein leakage(52).
These data suggest that RPE affected by the CNV may still be viable in the early stages of the disease (Fig. 15).
These studies have also reported that areas with previously high levels of FAF may show decreased FAF six months later(36).
These changes may be secondary to photoreceptor loss, RPE atrophy, replacement of normal phenotypes of RPE cells with scar, and increased melanin deposition.
These findings may have therapeutic implications and clarify long-term visual prognosis.
For example, a person with an active CNV on fluorescein angiography, and normal FAF, may show a much better outcome than another with an abnormal basal FAF.
Data comparing FAF findings in occult and classic CNV are limited.
Spital et al. reported that classic CNV usually shows more focal areas of decreased FAF than occult CNV(34).
These findings have been confirmed by McBain et al.(54) who guessed that low FAF at the site of the CNV are related to blocked fluorescence induced by the presence of CNV in the subretinal space, rather than to severe damage to the RPE.
A continuous preserved autofluorescence pattern was observed in the central macula in most of the cases.
These findings suggest that neovascular complexes, regardless if classic or occult, would be external to the RPE in most cases.
Additional studies with a higher number of patients and longer monitoring are required to verify with these changes in patients with CNV (Fig. 17).
Pigment epitelial detachment (PED)
PED can show different FAF patterns(11,16,31,55). FAF can provide complementary information to that of fundus colour photograph and fluorescein angiography.
In most of the cases, a moderately and diffusely increased FAF can be found, surrounded by a clearly defined ring with decreased fluorescence(34,35,55).
Occasionally, intermediate or even decreased FAF can be found that may not correspond to the atrophic RPE or to fibrovascular scars.
These changes in the FAF could correspond to different stages in the PED evolution(16).
The findings in FAF should be compared with those in fluorescein angiography.
We should bear in mind that areas with increased FAF do not always correspond to increased or decreased LF.
RPE tears usually occur in association with pigment epithelial detachments (PED) in patients with neovascular AMD, either spontaneously or following therapy(51).
FAF imaging reveals absence of autofluorescence in the area denuded from RPE.
These areas are clearly identifiable by their very low signal, whereas a heterogeneous FAF signal is seen in the area where the RPE is rolled.
Therefore, the exact location of the tear can be delineated in most cases.
FAF imaging is a very good tool to diagnose RPE tears(34,56).
The appearance of disciform scars in FAF imaging depends on their duration and evolution(34,36).
Disciform scars may show different variations and alterations of FAF signal.
A decreased signal is typically observed in scarred and fibrotic areas.
It has been reported that approximately 50% of the disciform scars may be surrounded by a rim of increased FAF(34,36).