2.1. Fluorescein Angiography & Indocyanine green angiography
Angiography is an intravenous dye-based imaging technique to study the circulation of the choroid and retina where, once dye is administered, a time-based sequence of fundus images is digitally recorded. Fluorescein angiography utilizes sodium fluorescein dye to illuminate the retina at a peak wavelength of 490 nm (blue) and then photographically record the excited fluorescent 530 nm (green) light that is emitted[
30] . Indocyanine Green (ICG) angiography utilizes ICG, a molecule that is 98% protein bound, and therefore remains in the fenestrated choriocapillaris longer and leaks less relative to fluorescein dye. It is sometimes more useful than fluorescein dye to study choroidal diseases such as nAMD[
31]
As imaging technologies evolve, their diagnostic and monitoring applications for AMD and MA have expanded. The course of MA in late AMD is characterized by the development of atrophic areas that enlarge continuously over time with cell death of the RPE, the outer neurosensory retina, and the choriocapillaris[
32] . In fundus fluorescein angiography (FA), atrophic patches appear as well-defined, hyperfluorescent areas because of enhanced visualization of the normal choroidal fluorescence caused by the loss of RPE cells (window defect), which would normally diminish the transmission of fluorescein fluorescence. Compared to fundus photography, this demarcated hyperfluorescent signal provides a sharper contrast between the atrophic and the surrounding non-atrophic areas. However, other pathologic findings in dry or wet AMD, such as drusen, pigmentary changes, fibrotic tissue, or neovascularization, may also result in an increased fluorescence signal or progressive dye leakage and therefore obscure the boundary demarcation of atrophy[
33,
34] .In indocyanine green angiography (ICG-A), atrophic patches appear as discrete hypofluorescent areas with loss of background fluorescence owing to small and medium vessel choriocapillaris atrophy[
35] .However, the large, deep choroidal vessels may still be visible and interfere with the outline of the area of atrophy and cause more difficulty in exact and reliable delineation. While ICG-A is useful in the differential diagnosis of polypoidal choroidal vasculopathy (PCV), chronic central serous chorioretinopathy (CSC), and retinal angiomatous proliferation (RAP), which are often misdiagnosed as nAMD[
36] , ICG-A has a negligible role for the identification of atrophy in AMD. In addition, both ICG-A and FA are invasive procedures that carry the risk of local infiltration, extravasation at the injection site, as well as the risk of an allergic reaction, which can be rare but severe and life-threatening, to the intravenously administered dye[
37] . FA is therefore recommended for the detection, classification, and quantification of NV but not atrophic changes; ICG-A can be used to distinguish other disease entities that cause atrophy[
38].
2.2. Fundus Autofluorescence
Fundus autofluorescence (FAF) is a non-invasive method, that provides rapid, noninvasive, high-contrast retinal images that are particularly useful for detecting atrophic areas, especially for better atrophic lesion boundary discrimination compared with color fundus photography[
39,
40] . FAF utilizes the fluorescent properties of lipofuscin, a byproduct of lysosomal breakdown of photoreceptor outer segments within the RPE cell. When excited by an appropriate light source, the bisretinoid components of lipofuscin absorb blue light with a peak excitation wavelength of approximately 470 nm and emit yellow-green light with a peak wavelength of 600 nm. A detector is then used to record the emissions signals as they are emitted. An FAF image, then, is a density map of lipofuscin where the brighter “hyperflourscent” areas represent areas of increased lipofuscin density and darker “hypflouresent” areas represent areas of decreased lipofuscin density[
41,
42]
One of the hallmarks of early and intermediate AMD is macular drusen[
43], which form with RPE aging. Drusen are composed of lipofuscin containing dense lipids, carbohydrates, zinc, and proteins, including apolipoprotein B and E, as well as components of the complement system[
44]. Recent grading systems, including the Age-Related Eye Disease Study (AREDS) and the Beckman Initiative for Macular Research Classification Committee, have classified drusen based on drusen type and size to associate drusen regression with or without RPE atrophy in NV or GA of late AMD[
45,
46,
47]
A recent study classified drusen-associated atrophy stages based on FAF and histological findings in eyes with late AMD[
48] . In stage 2, the earliest stage with detectable findings, FAF exhibited uniform hyperautofluorescence, indicating photoreceptor photopigment loss, whereas hypoautofluorescence in stages 3 and 4 corresponded to varying degrees of RPE atrophy. The FAF appearance is initially hyperfluorescent (stage 2), followed by a hypoautofluorescent center surrounded by hyperautofluorescent borders when associated with focal areas of RPE atrophy (stage 3), and hypoautofluorescent lesions with complete RPE loss (stage 4)[
48] . As the disease progresses through stages, the proportion of lipid within the drusen decreases relative to the proportion of calcification, with 80% of the drusen being refractile at the advanced stage 4. The refractile drusen appear as yellowish-white, glistening lesions and are associated with an increased risk of developing late AMD; however, they are undetectable on FAF alone[
49]
Of note, it was reported that cuticular drusen are strongly associated with late AMD[
50] .Eyes with the cuticular drusen can develop NV or acquired vitelliform lesions (AVL)[
51] which may regress to GA or RPE atrophy[
52] . In longitudinal studies, GA developed in 19.0% of eyes with cuticular drusen over a mean follow-up period of 40±18months, whereas GA developed in 28.4% and NV in 12.5% over a 5-year follow-up period[
53] . The cuticular drusen apex is steep and is where the atrophic RPE is located. FAF is an effective method to detect cuticular drusen with the display of numerous hypo-autofluorescences corresponding to the apex of the cuticle drusen with hyperautofluorescent rims. However, some FAF imaging cameras with different excitation wavelengths may not visualize these drusen[
54]
Studies showed that reticular pseudodrusen (subretinal dursenoid deposits) are highly associated with late AMD, such as GA, Type 3 macular NV, and drusenoid PED[
55,
56,
57] .Soft drusen are located beneath the RPE whereas reticular pseudodrusen are found on the surface of RPE[
58] . Studies classify reticular pseudodrusen into 3 types[
59,
60] in which the ribbon/reticular type is likely to progress to advanced AMD, including GA and Type 3 macular NV[
61,
62,
63] . Similar to cuticular drusen, eyes with the reticular pseudodrusen can develop NV or regress to GA or outer retina atrophy with focal photoreceptors loss and choroidal thinning[
64] .FAF may demonstrate a reticular pattern in eyes with reticular pseudodrusen; however, studies indicate that FAF is not the most specific method for detecting reticular pseudodrusen[
65]
Assessing the risk of late AMD depends on stratifying the types of drusenoid deposits and RPE abnormalities; and requires correctly evaluating imaging characteristics. The high-contrast differentiation of atrophic versus nonatrophic areas by FAF is a reliable image quantification of lesion area[
66] . Currently, conventional blue light excitation with excitation wavelength of 488 nm is the most popularly used mode for FAF imaging. However, macular pigment blocks blue light, resulting in a relatively diminished signal intensity at the fovea, which appears as a zone of hypofluorescence[
67] .Therefore, blue-light FAF may result in an overestimation of atrophic patch size and be mistaken for central atrophy involvement. The relative hypofluorescence of the fovea could mask an atrophic area, making it challenging to identify central minimal atrophic changes or adjacent paracentral atrophic margins[
68] . The quality of blue FAF signal may also be affected by pupil size or media opacity such as cataracts or vitreous opacity. FAF imaging systems include confocal scanning laser ophthalmoscopy (cSLO) systems and flash fundus camera-based systems. cSLO has FAF imaging with two excitation wavelengths (488 nm and 514 nm), while fundus camera autofluorescence relies on excitation wavelengths in the green to orange range (510-610 nm). One study reported that green-light FAF images (514 nm) are superior to blue autofutofluorescence (488 nm) for the evaluation of small central GA lesion size[
67] . Although the measurement of the atrophic lesions size in current clinical studies depends mainly on blue-light FAF, green-light FAF appears to be a more accurate, and a potentially important evaluation tool for central MA progression in future studies.
In certain phenotypic variants of GA, the loss of contrast between intact and atrophic RPE can have an altered FAF appearance, which differs from the markedly hypo-autofluorescent images in other forms of GA[
69] . In eyes with hemorrhagic nAMD or late nAMD with MA, the FAF signal may be reduced, and it is difficult to distinguish between atrophy and areas of fibrosis using FAF alone[
60] . Recently, blue-light FAF has been utilized in conjunction with near-infrared reflectance (NIR), which is unaffected by luteal pigment and enhances foveal evaluation. NIR is characterized by a long excitation wavelength (820 nm diode laser) [
70] that avoids the absorption of a shorter wavelength of light (480 nm) by melanin and lipofuscin granules at the RPE level, thereby allowing visualization of the retina and choroid[
71,
72] .Specifically, NIR reveals sub-RPE lesions effectively. Refractile drusen, for instance, are highly reflective, seen as glistening dots using NIR that are undetectable using FAF[
73] . Studies have reported that NIR has a very high sensitivity for detecting reticular pseudodrusen[
74,
75,
76,
77] .However, systematic validation studies for NIR alone in the detection of atrophic AMD are still lacking. Hence, FAF combined with other diagnostic modalities such as NIR may improve visibility of the obscured atrophic demarcated areas compared to when using FAF alone. Furthermore, widefield imaging devices can be used for the acquisition of FAF and both FA and ICG-A. Widefield imaging with a field of view that exceeds100 degrees and extends up to 150 degrees enables visualization of larger areas of the retina. Widefield imaging can monitor peripheral abnormalities to provide a more complete understanding of AMD[
78]
2.3. Optical Coherence Tomography
OCT is a noninvasive imaging modality that utilizes transversely scanned short coherence length light with interferometry to generate 2 dimensional and 3 dimensional cross-sectional maps of the retina and choroid with micrometer level resolution [
79] . While FAF is valuable for quantifying RPE loss in MA, it does not discern non-RPE layer changes[
68] .The classic definition of atrophy has been revised to incorporate changes in the outer retinal layers based on optical coherence tomography (OCT) findings[
69] . A classification system and criteria for OCT-defined atrophy associated with AMD has been proposed by The International Classification of Atrophy Meetings (CAM). According to the CAM study group, the OCT finding of atrophy undergoes an evolution of 4 different stages[
80] : (1) incomplete outer retinal atrophy, (2) complete outer retinal atrophy, (3) incomplete RPE and outer retinal atrophy (iRORA), and (4) complete RPE and outer retinal atrophy (cRORA). Of note is that these terms apply to atrophy in both non-neovascular (dry) and neovascular (wet) forms of AMD[
69] .The correlation between FAF changes and the four distinct atrophy categories is currently unknown. The severity to which hypoautofluorescence in FAF correlates with a single category of OCT-defined atrophy requires further investigation.
It is crucial that high resolution 3-dimensional OCT help identify the early phase of the atrophic process prior to lesion detection in 2-dimensional FAF [
69,
70,
71,
72,
73] The high axial resolution of Fourier-domain OCT devices, including spectral-domain OCT (SD-OCT) and swept-source OCT (SS-OCT), allows for the study of atrophy to quantify specific retinal layer loss. The wide application of SD-OCT has revolutionized the diagnosis and management of nAMD as it can provide assessment of risk and treatment prognosis, including the need for repeated anti-VEGF injections and other therapeutic intervention[
81] . Currently OCT has evolved into an effective imaging modality for evaluating early AMD changes. High-resolution OCT detects the presence of drusen and pigmentary changes in the early stages of AMD[
82,
83,
84], but SD-OCT provides important information regarding changes in retinal layers such as the outer plexiform layer (OPL), inner nuclear layer (INL), external limiting membrane (ELM), and ellipsoid zone (EZ). Unlike previously reported non-unique risk factors for the development of atrophy, such as hyperreflective foci and drusen characteristics including heterogeneous internal reflectivity and maximum drusen height and choroidal thickness beneath the drusen[
85,
86] , SD-OCT may detect unique early features such as the subsidence of the OPL and INL and a hyporeflective wedge-shaped band within the limits of the OPL, that represent significant risk and are present prior to development of drusen-associated atrophy[
87] . In addition, SD-OCT can detect early morphological changes before conventional diagnostic instruments can. For instance, in one study, SD-OCT showed that 2.9% of eyes with drusen-associated atrophy were already present in patients classified as having intermediate AMD on color fundus photography[
47,
87] . In another study, the pathological SD-OCT features occurred approximately one year prior to the development of definitive drusen-associated atrophy[
87] .This may enable treatment to be considered at an earlier time point in order to halt the progression of atrophy[
88,
89,
90] ,before late atrophic changes are detectable by conventional diagnostic methods.
A consensus was reached on the descriptions of imaging characteristics associated with atrophy or atrophy progression risk in eyes with AMD[
80] .These OCT features at risk for atrophy include intraretinal hyperreflective foci, extracellular deposits (soft drusen, drusen with hyporeflective cores, cuticular drusen, drusenoid PED, and subretinal drusenoid deposits), hyperreflective crystalline deposits in the sub-RPE basal lamina (BL) space and acquired vitelliform lesions[
74,
91,
92,
93,
94,
95,
96,
97] .As drusen regress, the overlying retinal layers progress to atrophy that can be detected by OCT imaging. Outer retinal atrophy features included INL and OPL subsidence, ELM descent, a hyporeflective wedge-shaped band within the Henle fiber layer, often accompanied by RPE disturbance and increased signal hypertransmission into the choroid, and ELM and EZ disruption[
98,
99,
100,
101] .For iRORA to be present, three OCT features, including photoreceptor degeneration, RPE attenuation or disruption, and increased signal transmission into the choroid, are required to be present[
102] . However, a minimum size limit for iRORA was not proposed. The study further reported that iRORA will progress and develop into cRORA over a variable time period ranging from months to years[
102] . A model was then developed to estimate future potential atrophy growth regions and identify predictive biomarkers. The most predictive SD-OCT biomarkers were thickness loss of bands, reflectivity of bands, thickness of reticular pseudodrusen, GA projection image, increased minimum retinal intensity map, and GA eccentricity, based on quantitative characteristics of GA[
103]
The anatomical correlations of the individual bands identified by an SD-OCT line scan are well established[
104] .The distance interval between scans must be small enough to avoid missing pathologic characteristics such as drusen, reticular pseudodrusen, and pigment migration into the inner retina. Scanning with a spacing of 125–250 μm is suggested for the detection of reticular pseudodrusen, which indicate rapid atrophy progression, and the volume rendering of outer retinal tubulations[
105,
106] .However, a less density scan is typically preferred in longitudinal, large-scale clinical trials as a trade-off to achieve shorter acquisition time[
107] . New advances in OCT imaging include SS-OCT, which offers faster scanning rates and a larger scan area[
108] , and enhanced depth OCT (ED-OCT), which employs enhanced depth imaging acquisition techniques to enable greater tissue penetration in the axial direction and the visualization of more choroidal details[
109] . Currently, detection of relevant clinical findings involves a volume scan by SD-OCT, or SS-OCT, but can be facilitated by screening with ED-OCT imaging to better detect the choroidal hyper transmission signal.
2.5. Optical Coherence Tomography Angiography
Imaging capable of providing appropriate visibility of the choriocapillaris and choroid has improved our understanding of atrophic and nAMD. While FA allows visualization of the retinal vasculature but not the choriocapillaris, ICG-A has not been widely utilized for choriocapillaris visualization in AMD due to its lack of depth resolution and inability to differentiate between choriocapillary blood flow and that of deeper choroidal vasculature [
115,
116,
117] . In contrast, optical coherence tomography angiography (OCTA) allows depth-resolved imaging of the retinal, choriocapillarial, and choroidal vasculatures. OCTA generates three-dimensional images of vasculature but without dye injection. Repeated imaging of stationary tissue with OCTA produces a series of identical B-scans; when there is motion due to blood flow, the repeated B-scans will alter, and the changes can be quantified[
118,
119,
120,
121] .Unlike dye-based angiography, such as FA or ICG-A, which is time-consuming and has a limited imaging window after injection, OCTA is quick and can be administered at any time during each patient visit. Recent OCTA studies demonstrated choriocapillaris loss across a spectrum of AMD phenotypes, including soft drusen, reticular pseudodrusen[
122,
123,
124,
125] ,and CNV[
126] .OCTA also allows for the evaluation of choroidal layers within and around atrophic lesions. Some studies found that the area surrounding the GA margin has greater choriocapillaris flow loss than the area of RPE atrophy or GA[
127] , indicating that choriocapillaris degeneration may occur prior to the development of GA and may be a prognostic factor for atrophic progression[
128,
129,
130,
131] .However, there are conflicting findings that choriocapillaris loss was linearly related to or less than RPE loss in GA[
132] , leading to the conclusion that the RPE appeared to be the primary target in GA[
133,
134] .In the GA region, it may be difficult to distinguish choriocapillaris flow impairment from atrophy due to OCTA's lower limit limitation in detecting slow blood flow. Increasing the interscan time can increase the sensitivity of OCTA to slow flows, but it also increases eye motion artifact noise[
135] . Hence, both the sensitivity to slow flow and the potential artifacts must be considered when interpreting OCTA data[
132] .In addition, OCTA limitations include acquisition time and field when used with conventional OCT. Therefore, dense, high-quality SD OCT or SS-OCT scans are required to obtain reliable OCTA results. The Consensus on Atrophy (CAM) study group recommended OCTA may be optionally included in studies on non-neovascular and neovascular AMD for exploratory purposes[
38] .