## Purpose

To evaluate the concordance between pseudodrusen as manifested by subretinal drusenoid deposits and large choroidal blood vessels using stereological analysis of spectral-domain optical coherence tomography (SD OCT) images.

## Design

Retrospective, observational case series.

## Methods

The SD OCT images of 31 consecutive patients with the clinical appearance of pseudodrusen from a private-referral retinal clinic were retrospectively reviewed. A grid of 19 evenly spaced vertical lines was randomly superimposed on each SD OCT image using ImageJ to perform systematic uniform random sampling. The main outcome measure was the likelihood of association between subretinal drusenoid deposits and large choroidal vessels.

## Results

Uniform random systematic sampling of 589 samples found the proportion of geometric probes intersecting subretinal drusenoid deposits to be 0.28, large choroidal vessel 0.65, and both 0.19. This value was nearly identical to the product of the joint probabilities and was within the 95% confidence interval (0.15–0.21) of the point estimate as calculated by the binomial theorem, indicating mutual independence. The subretinal drusenoid deposits were associated with neither large choroidal vessels nor the intervals in between.

## Conclusions

Our results demonstrate that there is no concordance between subretinal drusenoid deposits and large choroidal vessels or the stroma in between. As a consequence, hypotheses postulating that subretinal drusenoid deposits are associated with large choroidal vessels or the choroidal stromal spaces should be abandoned. Stereological techniques are powerful methods used in image evaluation in other fields of study and appear to have utility in analyzing OCT findings of the retina and choroid.

Pseudodrusen, first described clinically as a yellowish interlacing macular pattern better seen in blue light fundus photography in 1990, is a strong independent risk factor for late age-related macular degeneration (AMD). Many terms besides pseudodrusen have been proposed for this clinical entity, including reticular drusen, reticular pseudodrusen, or reticular macular disease. Zweifel and associates demonstrated that eyes with pseudodrusen have collections of material in the subretinal space, as seen using spectral-domain optical coherence tomography (SD OCT), that have the size and shape corresponding to the pseudodrusen seen in color fundus photographs, and suggested the term subretinal drusenoid deposits. This clinical observation supported previous laboratory observations made of mounds of subretinal material seen in autopsy eyes. The SD OCT characteristics of subretinal drusenoid deposits varied from diffuse granular hyperreflective material between the retinal pigment epithelium (RPE) and the ellipsoid zone to conical hyperreflective material sitting on the RPE and breaking through the ellipsoid zone.

There is controversy about pseudodrusen location and patterning. A hypothesis that there is a relationship between pseudodrusen and choroidal vessels first started when Arnold and associates postulated from the histopathologic analysis of 1 eye in which the retina was artifactually lost. The appearance of pseudodrusen was thought to be related to fibrous replacement of the choroidal stroma between a reduced number of choroidal vessels. However, the same group later reported the clinicopathologic correlation in a patient with pseudodrusen in a specimen in which the retina was not lost. This patient showed subretinal drusenoid deposits; the authors withdrew their suggestion that pseudodrusen appearance was related to fibrous replacement of the choroid. Since Arnold and associates first hypothesized that pseudodrusen resulted from fibrous replacement of the choroidal stroma, multiple imaging studies also proposed that there was some kind of concordance between pseudodrusen and the choroidal vasculature, stating that pseudodrusen were large vessels in the choroid (Haans R, et al. IOVS 2011; ARVO E-Abstract 1782); followed the edges of large vessels ; or followed, but were not located directly over, large choroidal vessels, the stroma between vessels, or the choriocapillaris. Although these seem quite disparate, a common element of all these theories is that the patterning of pseudodrusen is in some way controlled by or at least associated with the choroidal vasculature.

In all these imaging studies, the hypothesis that there was an association between subretinal drusenoid deposits and choroidal vessels was based on nonrandom selection and subjective assessment. Deriving the relationships, or lack thereof, between lesions and nearby structures can lead to better understanding of disease pathogenesis. As such, elucidating the exact relationship between pseudodrusen and the underlying choroidal vasculature is essential in understanding pseudodrusen, a recently recognized manifestation of AMD. In the present study, we used unbiased stereological evaluation to investigate whether or not there is co-localization between pseudodrusen as evidenced by subretinal drusenoid deposits on SD OCT imaging and large choroidal vessels. Stereology is an efficient method that uses geometric probes and inferential statistics to obtain unbiased estimates of quantitative data. Commonly used in microscopy and evaluation of neurobiologic specimens, it is employed here in an in vivo analysis of SD OCT of the eye.

## Methods

## Subjects

This retrospective study was approved by the Western Institutional Review Board and complied with the Health Insurance Portability and Accountability Act of 1996. The SD OCT images showing the detail of the outer retina and the scleral-choroidal interface of 31 consecutive patients with pseudodrusen clearly identified as subretinal drusenoid deposits stage 2 and 3 in SD OCT from a private retinal referral clinic were retrospectively reviewed.

## Imaging Definitions

The subretinal drusenoid deposits were defined as isolated mounds of hyperreflective subretinal material in SD OCT scans sufficient to alter the contour of the ellipsoid zone or conical hyperreflective material breaking through the ellipsoid zone, corresponding to subretinal drusenoid deposits stage 2 and 3, respectively, as previously defined. The large choroidal blood vessels were identified by a hyperreflective wall and dark center located in the outermost choroidal layer, close to the choroid-sclera junction.

## Image Acquisition and Processing

The SD OCT scans were obtained with the Heidelberg Spectralis (version 1.6.1; Heidelberg Engineering, Heidelberg, Germany), viewed with the Spectralis Viewing Module (version 5.7.0.1; Heidelberg Engineering). The imaging protocols were horizontal-line scan with varied scan size from 15 × 5 degrees to 30 × 20 degrees. The brightness and contrast of selected SD OCT images were manually adjusted with the Spectralis Viewing Module and Adobe Photoshop CS6 Extended (version 13.0.1 x64; Adobe Systems Inc, San Jose, California, USA) to optimize the delineation of subretinal drusenoid deposits and large choroidal vessels. We exported a SD OCT image containing subretinal drusenoid deposits of each subject from the Spectralis Viewing Module as TIFF (Tagged Image File Format; Adobe Systems, Inc) files for further analysis. The SD OCT images were imported to ImageJ (Java image processing program developed at the National Institutes of Health, Bethesda, Maryland, USA) and were straightened using the segmented line function to line along the curved RPE and using the command Edit > Selection > Straighten. Then a grid of 19 evenly spaced vertical lines was randomly superimposed on each SD OCT image to perform uniform random systematic sampling using the grid plug-in (written by Wayne Rasband; available from the National Institutes of Health at http://rsb.info.nih.gov/ij/plugins/grid.html ).

## Derivation of Stereology Method

A vertical line probe drawn at random perpendicularly through an OCT section will intersect a structure with a probability determined by the width of the structure in the OCT scan divided by the total width of the OCT scan. This would apply to both subretinal drusenoid deposits and large choroidal vessels. Taken in aggregate, the number of structures times their respective widths divided by the length of the scan defines the length fraction of that aggregate of structures.

Let V denote the event of the line intersecting a large choroidal vessel and ˉVV¯¯¯

V ¯

that it does not. Let S denote the event of the line intersecting a collection of subretinal drusenoid deposits and ˉS

S ¯

that it does not. The proportion of lines drawn at random that would intersect a large choroidal vessel, P(V), is equal to LV/(LV + L ˉV

V ¯

), where LV is the length fraction of the choroidal vessels in the section and L ˉV

V ¯

is the length fraction of where the large choroidal vessels are not, or in other words the intervening choroidal stroma. In a similar fashion P(S), the proportion of the probe lines crossing a subretinal drusenoid deposit, is equal to LS/(LS + L ˉS

S ¯

), where LS is the length fraction of subretinal drusenoid deposits in the section and L ˉS

S ¯

is the length fraction of the normal intervening sections of outer retina. If we assume they are independent, P(V and S) = P(V)P(S).

By setting up a grid of vertical lines through the section, the number of times a line crosses a vessel only, subretinal drusenoid deposits only, neither, or both will be counted. The measured probability of the line going through both the vessel and a subretinal drusenoid deposits is p, which is an estimator for the true proportion P. This process is repeated for each patient; the placement of the lines, although periodically spaced, is offset by a random distance each time. This causes a systematic uniform random sampling. From these measurements the probabilities of crossing a vessel, pV, and subretinal drusenoid deposit, pS, will be calculated. The joint probability, p, as per the null hypothesis, should be pVpS. If this probability is not near zero or 1 then the 95% confidence interval is given by the approximation from the normal distribution