Calcified Drusen Prevent the Detection of Underlying Choriocapillaris Using Swept-Source Optical Coherence Tomography Angiography

Abstract

Purpose: In age-related macular degeneration (AMD), choriocapillaris flow deficits (CCFDs) under soft drusen can be measured using established compensation strategies. This study investigated whether CCFDs can be quantified under calcified drusen (CaD).

Methods: CCFDs were measured in normal eyes (n = 30) and AMD eyes with soft drusen (n = 30) or CaD (n = 30). CCFD density masks were generated to highlight regions with higher CCFDs. Masks were also generated for soft drusen and CaD based on both structural en face OCT images and corresponding B-scans. Dice similarity coefficients were calculated between the CCFD density masks and both the soft drusen and CaD masks. A phantom experiment was conducted to simulate the impact of light scattering that arises from CaD.

Results: Area measurements of CCFDs were highly correlated with those of CaD but not soft drusen, suggesting an association between CaD and underlying CCFDs. However, unlike soft drusen, the detected optical coherence tomography (OCT) signals underlying CaD did not arise from the defined CC layer but were artifacts caused by the multiple scattering property of CaD. Phantom experiments showed that the presence of highly scattering material similar to the contents of CaD caused an artifactual scattering tail that falsely generated a signal in the CC structural layer but the underlying flow could not be detected. Similarly, CaD also caused an artifactual scattering tail and prevented the penetration of light into the choroid, resulting in en face hypotransmission defects and an inability to detect blood flow within the choriocapillaris. Upon resolution of the CaD, the CC perfusion became detectable.

Conclusions: The high scattering property of CaD leads to a scattering tail under these drusen that gives the illusion of a quantifiable optical coherence tomography angiography signal, but this signal does not contain the angiographic information required to assess CCFDs. For this reason, CCFDs cannot be reliably measured under CaD, and CaD must be identified and excluded from macular CCFD measurements.

Age-related macular degeneration (AMD) affects the central part of the retina known as the macula and is a leading cause of irreversible blindness among the elderly worldwide.¹ Optical coherence tomography (OCT) imaging provides non-invasive and rapid high-resolution three-dimensional (3D) images of the macular anatomy, and it has become the modality of choice to stage, treat, and monitor disease progression in AMD.^2–4 These images offer valuable insights into the progressive microstructural and microvascular changes that occur in AMD. The widespread use of OCT in ophthalmology has facilitated the identification of OCT biomarkers associated with the progression of changes from early to late AMD. These biomarkers include the area and volume of typical soft drusen,⁵ the area and volume of calcified drusen (CaD),⁶ subretinal drusenoid deposits,⁷ hyperreflective foci,^8,9 basal laminar deposits,¹⁰ the presence of persistent choroidal hypertransmission defects (hyperTDs),¹¹ lesions consistent with complete retinal pigment epithelium (RPE) and outer retinal atrophy,^12,13 and the presence of choriocapillaris flow deficits (CCFDs).^14,15 In particular, the extent of CCFDs, detected using OCT angiography (OCTA), has been associated with drusen, the growth of choroidal hyperTDs, a decrease in low-luminance visual acuity (LLVA), an increase in LLVA deficits, and delayed dark adaptation.^16–18 The local changes in CCFDs have shown a correlation with the growth rate of geographic atrophy.¹⁶ Thus, understanding the nuances of local variations in choriocapillaris flow can provide insight into disease progression and suggest potential therapeutic targets, especially in early stages or surrounding areas not yet affected by atrophy. 

Drusen are formed due to the extracellular deposits that accumulate between the basal lamina of the RPE and the inner collagenous layer of Bruch's membrane (BM).^19–22 The presence of two different types of calcifications (micron-sized spherules and calcified nodules) has been reported in the literature.^23–25 Although their differential impacts on disease progression and imaging have not been fully explored, the micron-sized spherules are found ubiquitously in all types of drusen, including soft drusen and basal linear deposits. These spherules are involved in drusen formation and growth. 

It is hypothesized that the presence of hundreds of these spherules within soft drusen, along with lipids and proteins, would attenuate but not completely block the OCT signals. The attenuated signals under these drusen can be detected in the choroidal structural layers. The attenuated signals under soft drusen tend to be mild and can be compensated using a strategy in which the inverted structural signal from the CC layer is used to enhance detection of the OCTA CC flow signal.^19,26 The underlying assumption when using this strategy is that, if a detectable attenuated signal is present, the compensation strategy can recover the diminished angiographic signal. This strategy has been validated in eyes where the drusen volume resolved spontaneously and the compensated CCFD measurements under drusen prior to drusen resolution were found to be comparable to the compensated measurements obtained after the drusen had resolved.²⁶ However, when the attenuation of the signal is complete, we refer to this completely attenuated signal as a choroidal hypotransmission defect (hypoTD),²⁷ and the compensation strategy used for typical soft drusen does not work with the CaD that cause hypoTDs. 

The presence of CaD, also known as refractile drusen on color fundus imaging or drusen with either hyperreflective or hyporeflective cores on OCT imaging, has been shown to be a biomarker for disease progression.^28–30 CaD contain calcium phosphate spherules,⁶ leading to increased optical absorption and light scattering.^31,32 Consequently, the calcified deposits impede the OCT light from penetrating into the choroid, resulting in hypoTDs in the choroidal layer.²⁹ This process is similar to the one caused by increased pigmentation within the retina, known as hyperreflective foci, or along the RPE border.^9,33 However, although we observed complete hypoTDs in the CC layer and deeper choroidal layers beneath hyperpigmentation, CaD resulted in a complete hypoTD specifically within the deeper choroidal layers (e.g., subRPE slab) but not in the CC layer. Additionally, we observed an absence of OCTA signal in the CC layer that could not be compensated, as the OCT signal in the CC structural layer did not appear to be attenuated. In order to explain this lack of attenuation in the CC structural layer, together with the complete attenuation of OCT signal in the deeper choroidal layers, we hypothesized that the highly scattering properties of the calcium phosphate spherules within the CaD, unlike the lipoprotein deposits within typical soft drusen, lead to the artifactual appearance of a structural signal under the CaD in the CC layer. This artifactual signal is caused by the multiple scattering of OCT light within CaD, giving rise to a tail-like false OCT signal beneath the CaD that does not carry meaningful information on CCFDs from the OCTA signal. Therefore, understanding the influence of CaD on CCFDs detection is important when using CC perfusion under drusen as an OCT biomarker for disease progression. 

The primary objective of this study was to investigate the impact of CaD on the detection of CCFD and to determine if CCFDs can be reliably measured under CaD. In this study, eyes were categorized into three groups: eyes with soft (or typical) drusen, eyes with CaD, and normal age-matched eyes. We proceeded to quantify the CCFD density by utilizing a heat map mask and outlining regions exhibiting abnormal concentrations of CCFDs or choriocapillaris flow impairment. We then evaluated the correlation between CCFDs and both typical drusen and CaD by using Dice similarity coefficients (DSCs) derived from the respective CCFD masks and lesion masks. Subsequently, we performed a phantom experiment to simulate the impact of the scattering particles within the CaD and to verify how the presence of CaD would obstruct light propagation into the choroid, while at the same time explaining a structural signal in the CC layer, combined with the absence of OCTA signal from the CC layer. 

Materials and Methods

This retrospective review of a prospective observational OCT imaging study was approved by the institutional review board of University of Miami Miller School of Medicine. The study was performed in accordance with the tenets of the Declaration of Helsinki and complied with the Health Insurance Portability and Accountability Act of 1996. Written informed consent was obtained from all participants before enrollment. Three groups of subjects, including eyes with CaD, eyes with soft drusen, and normal eyes, were identified from the natural history study and imaged from June 2016 to June 2022 with 3-month intervals over 9 months. Axial length measurements were performed using a noncontact biometry instrument (IOLMaster; Carl Zeiss Meditec, Dublin, CA, USA). Exclusion criteria included confounding ocular conditions such as an axial length ≤ 23 mm or ≥ 26 mm, glaucoma, or chorioretinal disorders other than dry AMD. 

Imaging Acquisition

All subjects were scanned using the PLEX Elite 9000 (Carl Zeiss Meditec). The swept-source OCTA (SS-OCTA) instrument employed a laser light source with a 1050-nm central wavelength and a 100-nm bandwidth, providing an axial resolution of approximately 5 µm and a lateral resolution of approximately 20 µm estimated at the retinal surface. All patients were imaged using the 6 × 6-mm scan pattern centered on the fovea. This SS-OCTA scan pattern consisted of 500 B-scans with 500 A-scans per B-scan, and each B-scan was repeated twice at each B-scan position, ensuring a uniform 12-µm spacing between fast and slow scans. The angiographic flow information was generated using the previously validated complex optical microangiographic (OMAG^c) algorithm.^2,34 Scans with a signal strength less than 7 based on the output of the instrument, as well as scans with significant motion artifacts, were excluded. 

Image Processing

Visualization and quantitation of the CC were performed using the 6 × 6-mm scans to compute the CCFDs within a 5-mm circle centered on the fovea. A schematic flow chart for the quantification process is shown in Figure 1. First, en face images were generated from a 3D SS-OCTA 6 × 6-mm scan that included an OCTA retinal microvascular image (Fig. 1A), a CC structural image (Fig. 1B), and a CC flow image (Fig. 1C), which are required in the compensation algorithm to adjust for any signal loss and projection artifacts due to the overlying anatomy, including the retina and the RPE/BM complex. After compensation, the CCFD map (Fig. 1F) was thresholded and binarized to quantify the CCFDs as previously described.^19,35,36 The retinal microvascular image (Fig. 1A) was generated by maximum projection of OCTA signals within a slab defined by the inner limiting membrane to 20 µm above the RPE layer (green segmentation lines shown in the corresponding B-scan in Fig. 1D). This image is used in the compensation algorithm to remove the overlying retinal projection artifacts in the CC flow image. The en face CC structural (Fig. 1B) and flow (Fig. 1C) images were generated by using a 16-µm-thick slab positioned 4 µm beneath the BM³⁷ using a maximum projection method. Before thresholding, any areas with hyperpigmentation^9,38 and persistent hyperTDs^27,39 were excluded from the compensated CC flow images in Figure 1F. The fuzzy C-means (FCM) thresholding algorithm was applied as previously described¹⁵ to produce binary CCFD maps, and CCFDs less than 24 µm in size were excluded from the final CC quantification (Fig. 1G). The CCFD percentage (CCFD%) was defined as the ratio of FD areas (bright areas in Fig. 1G) to the total quantifiable area. CCFD% values within a 5-mm circle centered at the fovea were also computed. 

Figure 1.

 

A schematic flow chart illustrating the image processes to generate and quantify the CCFDs and to generate the CCFD masks and CaD masks for later comparison, taking an eye with both soft and CaD as an example. (A–C) En face images that were generated from the 3D OCTA scans, including en face OCTA retinal flow image (A), OCT CC structural image (B), and OCTA CC flow image (C). (D, E) Representative B-scan image located at the position marked as dashed yellow lines in A, B, and C, illustrating the segmentation lines to define the slab (bounded by green lines in D) to generate the en face retinal flow image shown in A and the slab (bounded by red lines in E) to generate the CC enface images shown in B and C, respectively. (F) The compensated CC flow image that was obtained by a validated compensation algorithm and followed by exclusion of any areas with hyperpigmentation and persistent hypertransmission defects. (G) Binary CCFD images showing flow deficit regions. (H) CCFD heat map generated from the binary CCFD flow image in G, showing the impaired regions of choriocapillaris. (I) Final CCFD density mask that was generated by applying a threshold from the heat map in H for comparison with the identified CaD. (J–M) The steps required for manual identification of CaD and to generate the corresponding masks. The CaD were identified and manually outlined (yellow outlines in L). The hypoTDs that appeared in the subRPE enface image (K) were confirmed by the corresponding B-scans (J). Finally, a mask was generated from the outlines showing the area occupied by the CaD (M). The generated CCFD mask (I) and CaD mask (M) were finally compared to compute a Dice similarity coefficient.

To facilitate the comparison between CCFDs and CaD, a binary mask showing impaired choriocapillaris perfusion regions was generated from the CCFD image (Fig. 1G). In doing so, we first generated a heat map (Fig. 1H) from the binary CCFD map in which the size of each individual CCFD was measured and a size threshold was applied to create the CCFD heat map using a Gaussian blur function, which represents the density of CCFDs.⁴⁰ A density threshold was then applied to the heat map to produce the final CCFD density mask (Fig. 1I), indicating the main area with impaired choriocapillaris perfusion. A grid search algorithm was used to determine the optimized size and density thresholds in this procedure,⁴¹ where a criterion of maximum correlation between the density mask and the CaD mask was used to constrain the search. This procedure resulted in the thresholds of 55 pixels for the size and 0.25 for the density, respectively. In the soft drusen group, the same thresholds for size and density were applied to obtain the CCFD density masks. 

Identification of Drusen

Area and volume measurements of typical soft drusen were generated using a validated algorithm known as the Advanced RPE Analysis algorithm (version 0.10), which is available on the Advanced Retinal Imaging Network website (Carl Zeiss Meditec).²⁸ The drusen volumes in the 3-mm and 5-mm circles centered on the fovea were obtained. CaD were identified as hypoTDs on the subRPE structural en face images generated using a slab from 64 µm to 400 µm below BM (Fig. 1K). CaD were confirmed using corresponding B-scans (Fig. 1J) as previously reported.¹ These CaD were manually outlined (Fig. 1L), and their area measurements within the outlined boundaries were calculated. Finally, a binary mask was generated showing the areas occupied by the CaD (Fig. 1M) for later computation of DSCs with a CCFD density mask as described in the last section (Fig. 1I). In the soft drusen group, the drusen masks were created from the drusen elevation maps generated from the SS-OCT Advanced RPE Analysis algorithm (version 0.10). 

Correlations Between CCFDs and Drusen

Correlations among CCFDs and both CaD and typical soft drusen were investigated (Fig. 2). To accomplish this, we evaluated the degree of overlap between the CCFD density masks (Figs. 1I, 2D) and the CaD masks (Figs. 1M, 2E), for which DSCs were computed and compared. The value of a DSC ranges from 0, indicating no spatial overlap between two sets of binary segmentation results, to 1, indicating a complete overlap.  

\begin{equation*}DSC = \frac{{2TP}}{{2TP + FP + FN}}\end{equation*}

where TP denotes true positive, FP denotes false positive, and FN denotes false negative. 

Figure 2.

 

Examples of CCFD density masks and drusen lesion masks for computing Dice similarity coefficients. (A–E) Same eye as in Figure 1 with CaD. (F–J) Eye with soft drusen. (A, F) Compensated CC flow images with the exclusion of any areas with hyperpigmentation and persistent hypertransmission defects. (B, G) Corresponding binary CCFD maps of A and F. (C, H) Corresponding CCFD heat maps of B and G. (D, I) CCFD density masks produced by using the density threshold on the CCFD heat maps in C and H. (E, J) Lesion masks generated by manual outlines of drusen and used to compare with CCFD density masks for the calculation of Dice similarity coefficients.

DSCs were also calculated between the CCFD density masks and the soft drusen masks (Figs. 2F–2J), where the soft drusen mask was created from the drusen elevation map generated from the SS-OCT Advanced RPE Analysis algorithm (version 0.10). Correlations among the square-root (sqrt) area of the CCFD density masks against the sqrt areas of CaD and soft drusen were analyzed and compared. 

Phantom Experiments

To investigate the influence of CaD on OCTA signals, a simplified three-layer phantom was designed and fabricated to simulate the optical properties of the retina tissue with the inclusion of highly scattering lesions to mimic CaD. The top layer is a transparent glass layer (to simulate the neural retina tissue that is relatively transparent to the light). To mimic the CaD, we deposited a number of droplets with a size of approximately 200 µm and a height of approximately 80 µm onto a glass slide. The droplets were made of solidified agar emulsion mixed with 1.5% titanium oxide (TiO₂), which gives an approximate optical attenuation (or scattering) coefficient of 3.0/mm. The choice of this droplet to mimic the CaD was empirical; however, its light scattering strength is similar to the optical attenuation property of calcium phosphate spherule aggregates.⁴² We then flipped the slide over so that the droplets were in direct contact with the thin empty channel designed to accommodate the flow of intralipid that served to mimic choriocapillaris/choroidal blood flow. If the slide had not been flipped over, then the glass slide would have been between the TiO₂ droplets and the flowing intralipid. Although the configuration of the TiO₂ droplets is opposite of the expected configuration of drusen, the construct allows for the direct approximation of the droplets to the choriocapillaris/choroidal flow, much as it exists in the back of the eye when drusen and the choriocapillaris are separated by a thin BM. A phantom with the structural configuration identical to the naturally occurring configuration of drusen and choriocapillaris requires advanced engineering capabilities that were not available to us. However, this current design adequately demonstrates the detection of an artifactual structural signal beneath the droplet that arises due to high light-scattering properties of the droplet, which obscures the detection of flow in the layer that would be analogous to the choriocapillaris layer that is the layer of interest. 

The bottom layer was constructed from a highly light-scattering solid base that was designed to mimic the properties of the sclera layer. Then, the dynamic flow was introduced through the infusion of a 5% intralipid solution into the flow channel (middle layer) to simulate the physiological conditions. Note that a 5% intralipid solution has scattering characteristics (with a reduced scattering coefficient of ∼1/mm) similar to blood. After its fabrication, the phantom underwent SS-OCT imaging using a home-built SS-OCT system with a central wavelength of 1050 nm. This setup facilitated the acquisition of 3D OCT and OCTA images. Notably, the home-built SS-OCT system demonstrated performance characteristics akin to clinical SS-OCT systems, ensuring the reliability and translatability of the acquired phantom data.⁴³ 

Statistical Analysis

Continuous data are described as mean ± SD. Independent sample t-tests and χ² tests were used to compare the results from two groups. Pearson's linear correlation compared the sqrt area of the CCFD density masks and the lesion masks. Statistical analyses were performed with SPSS Statistics 28 (IBM, Chicago, IL, USA) with P < 0.05 considered to be statistically significant. 

Results

A total of 90 eyes were selected, among which 30 eyes (29 patients) had CaD, 30 eyes (26 patients) had soft drusen, and 30 eyes (30 patients) served as normal controls. The Table shows that the eyes were well matched with respect to age and gender across the three groups. In addition, drusen volumes within the fovea-centered 3-mm and-5mm circles were matched between the CaD and soft drusen groups (both P > 0.05). 

Table.

 

Characteristics of Subjects and Eyes With CaD, Typical Soft Drusen Eyes, and Eyes With no Obvious Disease

 

Figure 3 displays representative cases of a normal eye (Figs. 3A–3F), a soft drusen eye (Figs. 3G–3L), and a CaD eye (Figs. 3M–3R) imaged using SS-OCTA scans to generate en face images, corresponding B-scans, drusen volume maps, and compensated CCFD binary maps. Compared with the normal eyes, the structural images from eyes with drusen showed signal attenuation under the soft drusen and complete signal loss beneath the hyperpigmentation (Figs. 3H, 3I, 3K). However, no signal loss was observed on the structural CC slab underlying the CaD (Figs. 3N–3P, white arrows), whereas the flow signal on the CC slab underlying the CaD was absent (Figs. 3Q, 3R, white arrows) 

Figure 3.

 

Examples of a normal eye (A–F), an eye with soft drusen (G–L), and an eye with CaD (M–R) imaged using SS-OCTA scans. (A, G, M) Drusen elevation maps generated from the SS-OCT Advanced RPE Analysis algorithm (version 0.10). The white 5-mm circles centered on the fovea indicate the areas that were used to measure the drusen volumes. (B, H, N) Corresponding OCT B-scans indicated by the dashed yellow lines on the subRPE structural images shown in C, I, and O. The segmentation boundaries used to generate the subRPE slabs are indicated by the yellow dashed lines in each panel. (C, I, O) En face subRPE structural images show evidence of choroidal hypoTDs, which correspond with the CaD (white arrows) in N, O, and Q and hyperpigmentation (yellow arrows) in the soft drusen eye in H, I, and K. (D, J, P) En face CC structural image from the normal eye shows a homogeneous gray area with the outlines of retinal vessels (D), and the eyes with drusen show shadowing from drusen and signal loss from hyperpigmentation (J, P). (E, K, Q) Compensated CC flow images were generated after excluding the areas of hyperpigmentation. (F, L, R) Binary CCFD images were computed to measure the CCFD% (white foci denote CCFDs). The 5-mm white circles centered on the fovea indicate the area that was used to analyze the CCFDs.

Figure 4 presents the OCT imaging characteristics of CaD. Enlarged images of the area underlying the CaD (Figs. 4C, 4H, 4I, 4K) show hypoTDs (white arrows) on the subRPE structural slab (Figs. 4C, 4H) and a hyperreflective scattering tail (orange arrows) on the CC structural slab and B-scans (Figs. 4E, 4I, 4K). These findings suggest an association among CaD, hypoTDs, and a hyperreflective scattering tail within the CC slab. 

Figure 4.

 

Illustration of OCT imaging characteristics of CaD. Shown is a representative example of an eye with CaD imaged with SS-OCTA. (A) Drusen volume map. (B) OCT structural image located at the position marked by the purple horizontal line in A. (C) Enlarged OCT structural image containing CaD. The orange arrow indicates the hyperreflective scattering tail, and the white arrow points to the hypoTD. (D) OCTA flow image at the location marked by the purple horizontal line in A. (E) En face subRPE image generated from a slab defined by segmentation boundaries at 64 to 400 µm below BM (yellow lines in K and L). (F) Enlarged image of the CaD marked in E. The white asterisk identifies the hypoTD under the CaD (pointed at by the white arrows in C and J). (G) En face CC OCT structural image generated from a slab defined by segmentation boundaries at 4 to 20 µm below BM (yellow lines in K and L). (H) Enlarged CC structural image beneath the CaD marked in F, where the orange star corresponds to the same CaD area identified with a white asterisk in F. (B, C, I, J) B-scans located at the positions marked by purple horizontal line in A and E to H without (B) and with (I) segmentation boundaries to define subRPE, and an enlarged view of the region containing CaD (J, C). (K, L) Corresponding B-scans with segmentation boundaries to define the CC slab. The white arrows in C and J point to the shadows seen within the subRPE slab causing the hypoTD appearance in E and F. The orange arrows in C and L point to the hyperreflective scattering tail seen within the CC slab that corresponds to the bright areas seen in G and H (marked by yellow stars). Taken together, these OCT images show that CaD is associated with hypoTDs on the subRPE slab while simultaneously giving rise to a detectable OCT signal within the CC slab that corresponds to a hyperreflective scattering tail that extends below the CaD and involves the CC slab.

Figure 5 shows the extent of CCFDs in the fovea-centered 5-mm circle in the three groups, specifically underlying the soft drusen and CaD. In the central 5-mm circle, the CCFD% underlying the CaD group exhibited a significant elevation compared with the soft drusen group (10.92% ± 1.88% vs. 9.39% ± 1.78%; P = 0.002). Furthermore, the CCFD% underlying the soft drusen group exceeded that in the normal control group (9.39% ± 1.78% vs. 7.82% ± 0.81%; P < 0.001). Notably, when quantified specifically under the drusen, the CCFD% measurements underlying the CaD were markedly higher (48.87% ± 15.94%) than those observed under the soft drusen (23.63% ± 6.97%; P < 0.001). 

Figure 5.

 

Comparison of the CCFDs among the three groups of eyes. (A) Comparison of normal eyes, soft drusen eyes, and CaD eyes within the fovea-centered 5-mm circles. (B) Comparison of CCFDs underlying the soft drusen and CaD.

Figure 6 shows the DSCs between the area masks of the different drusen types and the masks of the CCFDs obtained using the CCFD heat maps. The DSCs served as an important metric to provide the correlation between the shape of CaD with the distribution of CCFDs. In the context of CaD, the median DSC of the CCFD density masks and CaD masks was computed at 0.51, with a mean ± SD of 0.45 ± 0.27. In the presence of soft drusen, the DSCs of the CCFD density masks and soft drusen masks were significantly lower, with a median of 0.15 and a mean ± SD of 0.20 ± 0.18 (P < 0.001). This analysis emphasizes the intricate interplay between the CCFD density measurements and the presence of different drusen types. 

Figure 6.

 

Dice similarity coefficients of choriocapillaris flow deficit density masks associated with CaD and soft drusen masks.

Figure 7 depicts the phantom model (Fig. 7A) that simulated how a lesion, such as CaD with high light-scattering properties (Fig. 7A, green arrows), can create a scattering tail (Fig. 7B, white arrows) that would impact the detection of an angiographic flow signal underlying the lesion (Fig. 7C, white arrows). The complex scattering of the light within the lesion leads to an increased optical pathlength which appears on the OCT image as if the light is reflected from a depth beneath the lesion. Consequently, an artifactual tail is created beneath the lesion (Fig. 7B, white arrows). On the other hand, the highly scattering property of the lesion heavily attenuates the incident light, dramatically reducing the chance of that light penetrating the deeper layers. This attenuation leads to a loss of OCTA flow signal beneath the lesion, and the high light-scattering material makes it impossible to reliably detect useful OCT and OCTA signals from these underlying structures. This phenomenon is highlighted by the white arrow in Figure 7C, where the existing real flow below the scattering lesion³⁰ is missing even though a signal appears to be present based on the structural image (Fig. 7B). However, the flow on either side of the scattering lesion can be detected. The results of this phantom study suggest that the detection of blood flow beneath the highly scattering lesion and within the tail cannot be reliably detected. 

Figure 7.

 

Phantom experiment to confirm the scattering tail due to highly scattering deposits that simulate CaD, demonstrating the inability to detect OCTA flow signals beneath the scattering lesions. (A) Experimental phantom model consisting of three-layer layers where the upper layer was a transparent glass layer with droplets of highly scattering material deposited on a slide to simulate CaD. The slide was then inverted, and the lesions appeared on the bottom surface (green arrow) adjacent to the flowing intralipid. The droplet was made of solidified agar emulsion mixed with 1.5% TiO₂ particles. The middle layer contained an empty channel, identified by a yellow arrow, to facilitate the infusion of a scattering liquid to mimic blood flow (5% intralipid solution). The base of the phantom was constructed from high scattering materials to simulate sclera. (B, C) Representative cross-sectional OCT structural image (B) and OCTA flow image (C) scanned from the phantom. The green arrows indicate areas with highly scattering droplets (mimicking CaD), and the yellow arrows point to the flowing intralipid solution. Notably, the scattering tails (white arrows in B) are apparent beneath the droplet due to the increased scattering of light within the mass that increases the optical path length when detected by the OCT; however, in C, the OCTA signals were not generated within the scattering tails where the flow should be present (white arrows), and this flow was detected adjacent to the scattering tails (yellow arrows).

Figure 8 depicts the natural history of the transformation of a lesion from soft drusen to the appearance of CaD to the appearance of hyperTDs when imaged at 3-month intervals over 9 months. This case example illustrates the absence and then reappearance of CC flow when the calcified material is cleared. This case also demonstrates an intriguing outcome showing that the appearance of the hyperTDs was not associated with an increase in CCFDs. The fluctuations in CCFD% beneath the lesion are particularly noteworthy. An initial increase was observed, from 10.15% (Fig. 8D) to 15.34% (Fig. 8I), followed by a subsequent decrease to 11.80% (Fig. 8N).¹⁹ This temporal pattern shows that the CC perfusion improves even as the lesion evolves into hyperTDs. 

Figure 8.

 

Evolution of a normal druse to a calcified druse to a hyperTD with a loss of CC perfusion under the CaD but with the reappearance of CC perfusion once the calcified druse resolved and the hyperTD appeared. This representative case was imagined over 9 months. The yellow square identifies the region containing the calcified druse of interest. (A–E) Soft drusen stage. (F–J) CaD stage. (K–O) HyperTD stage. (A, F, K) The en face subRPE images over time; the location of the lesion is indicated by the yellow square. (B, G, L) The OCT B-scans along the green dashed line in A, F, and K highlighting the choroidal hypoTD resulting from the attenuation of light due to the calcified druse (F, G), followed by the hyperTD when the calcified druse resolved, resulting in attenuation or loss of the retinal pigment epithelium and outer retina (K, L). (C, H, M) Compensated CC flow images in which hyperpigmentation and other CaD outside the yellow square have been masked and excluded from the analysis. (D, I, N) The corresponding binary CCFD maps. (E, J, O) The corresponding CCFD density heat maps.

Discussion

The findings from this study highlight the impact of CaD on attempts to quantify CC perfusion. Although a structural signal from the CC layer can be detected under the CaD, this signal provides false depth information. Consequently, it cannot be used to extract OCTA information about blood perfusion because the signal within the scattering tail results from the highly scattering properties of the CaD themselves, lacking any useful detectable flow signal. We hypothesize that the highly light-scattering properties of the calcium phosphate spherules within the CaD result in increased optical pathlengths of the light originating from the CaD, thereby forming a scattering tail directly below it. It is crucial to note that that OCT signal represents the detection of the light based on its optical pathlength to provide depth-resolved information. Hence, the artifactual illusion of depth information results from the delayed reflectance of light back to the detector due to the increased optical pathlength of the light originating within the CaD, caused by its high light-scattering properties. Consequently, the signal under the CaD does not arise from the CC layer but rather from within the CaD, so this signal lacks any flow information. Thus, when using the CCFD algorithm to detect CC perfusion, a decrease in CC perfusion under CaD, reported as an increase in CCFDs, is an artifact due to the absence of both structure and angiographic information from beneath the CaD. The scattering tail from the CaD provides a false structural signal in the CC layer, whereas in the deeper choroidal layers the strong light attenuation due to the CaD causes a choroidal hypoTD that appears as a dark region on the en face subRPE structural slab. Unlike soft drusen, in which the OCT signal is mildly attenuated as it passes through the drusen and can be compensated for so that the CC perfusion can be quantitated,²⁶ there is no real signal under the CaD that can be compensated. Therefore, CC perfusion cannot be quantified under CaD using current imaging technology and compensation strategies. 

Our study evaluated 90 eyes, categorized into groups representing different drusen types and normal controls, and the observed variations in CCFDs among these groups demonstrated the impact of CaD on the detection of CCFDs. Using the DSCs and area analyses to quantitatively assess the relationship between CCFDs and CaD, we demonstrated a strong correlation between CCFD density and CaD, reinforcing the erroneous belief that the presence of CaD is associated with decreased CCFD measurements.⁴⁴ Our results demonstrate the need to interpret OCTA data in regions affected by CaD cautiously and to consider the influence of these structures on CCFD quantification. Although other studies have associated the decrease in CC perfusion with drusen as a likely cause of drusen formation and disease progression,^45-49 these studies must be viewed with extreme caution, as these groups did not meticulously exclude the full extent of CaD from their analyses. This concern about the relationship between CCFDs and disease progression is demonstrated by the case depicted in Figure 8. When the soft druse was first present, the CCFDs could be quantified; however, when the druse developed calcifications, the CCFD% increased, but this increased CCFD% returned to its initial state when the calcified material resolved and a hyperTD formed. The return of CC perfusion highlights the artifactual nature of the presumed decrease in CC perfusion under CaD. Although it may be possible to restore CC perfusion when the calcification resolves, the more likely explanation is that the presence of the calcified material prevents the detection of CC perfusion. Another observation was that we expected an increase in CCFD% when the hyperTD formed, but that is not what we found. Additional cases showing the evolution of hyperTDs should be studied, but this case suggests that the RPE changes that result in the formation of a hyperTD are not preceded by a significant decrease in CC perfusion.^14,50–53 

Our phantom experiment provided insight into the challenges posed by CaD-like structures in OCTA imaging. The experiment revealed that the presence of CaD introduces scattering tails that can obstruct the OCTA signals from underlying structures. The light attenuation due to high light-scattering material results in the absence of an OCTA flow signal beneath the lesion that cannot be compensated using the published strategy of inverting the decreased structural slab signal because the signal from the beneath the CaD is from the scattering tail and does not contain any structural or flow information. When the concentration of highly light-scattering material exceeds a certain threshold, the OCT and OCTA signals originating from structures underlying the lesion cannot be reliably detected because they are below the minimal OCTA detection range. This phenomenon aligns with our clinical observations, emphasizing the need for the identification of CaD and their exclusion from macular CC perfusion studies. Although current CC algorithms can compensate for attenuated signals, the highly scattering calcium phosphate spherules along with their scattering tails present an unsolved challenge that prevents the accurate detection of blood flow patterns under CaD. 

Limitations of our study include a small number of eyes in each group; however, considering the statistical significance of our findings, it is highly unlikely that our conclusions will change with additional eyes. Further studies with larger and more diverse cohorts would be useful to validate and extend our findings. Although the qualitative nature of the phantom model appears to be consistent with our clinical observations, it is certainly possible that the properties of CaD are not analogous to a 1.5% TiO₂ agar suspension. However, the phantom does unequivocally demonstrate the creation of a scattering tail and the impact of the scattering tail on the detection of an OCTA flow signal. We chose a simplified and cost-effective phantom experiment to illustrate the inability of OCTA to detect flow beneath a highly light-scattering nodule-like deposit. A more sophisticated phantom model, capable of mimicking the scenarios of CaD within pathological and physiological contexts, would be necessary if the aim is to investigate how the scattering signals originating from CaD could be used to determine the physical composition of the drusen. 

In summary, our findings underscore the necessity to carefully analyze and interpret CCFDs in eyes with CaD. The easiest way to identify the CaD is by examining the subRPE slab derived from a dense volumetric scan pattern of the macula. These CaD can be identified as choroidal hypoTDs and confirmed on corresponding B-scans to exclude areas of hyperpigmentation, which also cause choroidal hypoTDs. When quantifying CCFDs, it is essential to identify these regions and exclude these hypoTDs before compensating and thresholding the CCFDs to avoid introducing artifacts into the quantitation of CCFDs. The presence of CCFDs underlying CaD is not real but results from the increased light-scattering properties of CaD. To further investigate the importance of CC perfusion as an OCT biomarker of disease progression in AMD, we must emphasize the need to identify and exclude these CaD before establishing any temporal relationship between the appearance of CC perfusion deficits and the progression of disease. 

Acknowledgments

Supported in part by an unrestricted grant from the Research to Prevent Blindness, Carl Zeiss Meditec, and grants from the National Institutes of Health (P30EY014801, R01EY028753, and R01AG060942). The funding organizations had no role in the design or conduct of the present research. 

Disclosure: Y. Cheng, None; F. Hiya, None; J. Li, None; M. Shen, None; J. Liu, None; G. Herrera, None; A. Berni, None; R. Morin, None; J. Joseph, None; Q. Zhang, Carl Zeiss Meditec (E); G. Gregori, Carl Zeiss Meditec (P, R); P.J. Rosenfeld, Annexon (C), Apellis (C, F), Bayer Pharmaceuticals (C), Boehringer-Ingelheim (C), Carl Zeiss Meditec (P, R), Gyroscope Therapeutics (C, R), InflammX Therapeutics (C, F), Ocudyne (C, F), Regeneron Pharmaceuticals (C), Stealth BioTherapeutics (C, R), Unity Biotechnology (C), Valitor (C, F); R.K. Wang, Carl Zeiss Meditec (C, R), Colgate-Palmolive (R), Cyberdontics (C), Estee Lauder (R), Oregon Health and Science University (I), University of Washington (I)