**Purpose**:
To provide a consistent implementation of a retinal ganglion cell (RGC) displacement model proposed by Drasdo et al. for macular structure-function analysis, customizable by axial length (AL).

**Methods**:
The effect of axial length on the shape of the inner retina was measured on 235 optical coherence tomography (OCT) scans from healthy eyes, to provide evidence for geometric scaling of structures with eye size. Following this assumption, we applied the Drasdo model to map perimetric stimuli on the radially displaced RGCs using two different methods: Method 1 only displaced the center of the stimuli; Method 2 applied the displacement to every point on the edge of the stimuli. We compared the accuracy of the two methods by calculating, for each stimulus, the number of expected RGC receptive fields and the number RGCs calculated from the histology map, expected to be equivalent. The same calculation was repeated on RGC density maps derived from 28 OCT scans from 28 young healthy subjects (age < 40 years) to confirm our results on clinically available measurements.

**Results**:
The size of the retinal structures significantly increased with AL (*P* < 0.001) and was well predicted by geometric scaling. Method 1 systematically underestimated the RGC counts by as much as 60%. No bias was observed with Method 2.

**Conclusions**:
The Drasdo model can effectively account for AL assuming geometric scaling. Method 2 should be used for structure-function analyses.

**Translational Relevance**:
We developed a free web App in Shiny R to make our results available for researchers.

^{1}

^{–}

^{3}It is now recognized that the macula can be affected by glaucoma, even in the early stages of the disease process.

^{4}Loss and dysfunction of retinal ganglion cells (RGCs) in glaucoma is monitored using both structural and functional measurements. Structural assessment of the macular region can be done by spectral domain optical coherence tomography (SD-OCT), which provides volumetric measurements of various retinal layers. The layers of most interest for glaucoma are the retinal nerve fiber layer (RNFL), the ganglion cell layer (GCL), and the inner plexiform layer (IPL). Together, they form the inner retina. RGC loss from glaucoma causes thinning of the RNFL (which contains RGC axons), the GCL (which contains RGC cell bodies), and the IPL (which contains the RGC dendritic arbors).

^{4}

^{,}

^{5}

^{6}

^{,}

^{7}Models seeking to match structural and functional data to histology measurements of RGC density have been used to explore this relationship.

^{8}

^{–}

^{11}Recently, a method proposed by Raza and Hood

^{12}has been used to convert the GCL thickness into RGC density to investigate the relationship between RGC number and SAP sensitivity

^{13}

^{,}

^{14}in healthy subjects and glaucoma patients.

^{15}

^{–}

^{17}RGCs are connected to the corresponding photoreceptors via Henle's fibers, which have an oblique pathway in the parafoveal region. This displacement diminishes with eccentricity, becoming minor at around 10 visual degrees from the fovea.

^{15}

^{,}

^{18}Different numerical models, based on, or verified by, histologic measurements of Henle's fibers have been proposed to account for this displacement.

^{15}

^{,}

^{18}

^{–}

^{20}The most widely used of these models is the one proposed by Drasdo et al.

^{15}This model is valuable in the context of structure-function analyses because the displacement calculation requires equivalence between the cumulative number of RGC receptive fields (RGC-RF), estimated through psychophysical measurements, and the number of RGC bodies, estimated through histology.

^{16}Theoretically, when applied correctly, this model would allow a one-to-one correspondence between the number of RGCs, estimated from structural maps, and psychophysical measurements, estimated from SAP sensitivity (which depends on the number of RGC-RFs stimulated during the test).

^{15}is not straightforward. For example, Drasdo et al. only reported numerical calculations for the four principal meridians and the average displacement, in microns.

^{15}Therefore, a method to generalize to any arbitrary meridian has not been available. A second example is that the schematic eye used by Drasdo et al.

^{15}

^{,}

^{21}to convert visual degrees to millimeters of retina assumes a spherical shape for the retina of a certain radius. However, that radius is not the same as that assumed by Curcio and Allen

^{16}in their published histology map of RGC density. Moreover, the radius of the sphere should be adjusted for the axial length (AL), when this is available. However, Drasdo et al.

^{15}

^{,}

^{21}only provided average displacement values regardless of axial length. A third example is that, in many cases, a simple displacement of the stimulus centers was applied.

^{7}

^{,}

^{13}

^{,}

^{22}However, the displacement should be applied to the perimeter of the stimulus, so that different points of the stimulus edge are independently displaced radially outward according to the model. For example, in the parafoveal region, the stimulus edge nearer the fovea is displaced further than the stimulus edge further from the fovea.

^{14}

^{,}

^{23}This is especially important when RGC counts are involved, because small differences in the area used for calculations can result in large differences in the counts.

*AL*was measured using a Lenstar LS 900 Biometer (Haag-Streit AG, Switzerland). These scans were further screened by an ophthalmologist (GM) for pathologic changes of the inner retina. Seventeen scans were excluded because of poor quality that prevented a clear identification of the inner retinal layers or the Bruch's membrane within 15° from the fovea, 13 scans had local thinning that could be attributed either to glaucoma or local ischemia and four were excluded for vitreoretinal alterations. The segmentation of the retinal layers was checked and manually corrected where necessary leaving 265 scans for analysis. This dataset was used exclusively to extract metrics on the shape of the GCL profile. No thickness values were measured. The median [interquartile range] quality index (QI) was 30.6 [28.98, 32.26] dB.

*AL*was measured using an IOL-Master V3 A-scan (Zeiss Meditec, Dublin, CA, USA). The subjects had no known or detectable ocular disease and younger than 40 (range 23–37) years, to match the age range of the histological dataset collected by Curcio and Allen

^{16}(see next section). Descriptive statistics for the two datasets are given in Table 1. Best corrected visual acuity in Dataset 2 was 0.00 LogMar for all subjects and was not measured further. All data collections were performed in agreement with the declaration of Helsinki after explicit written consent from the participants. All scans were of good quality and none was excluded (QI = 26.47 [25.36–27.47] dB).

^{15}used the histology map provided by Curcio and Allen.

^{16}This reports the density of ganglion cells (cells/

*mm*

^{2}) obtained from six retinas of five healthy subjects, aged 27 to 37 years (range), for a retinal sphere with a radius of 11.459

*mm*. Details are reported in Appendix A.

^{21}and later used by Drasdo et al.

^{15}for their displacement model. We used numerical ray tracing through the schematic eye to calculate the correspondence between visual degrees and

*mm*, and solid visual degrees and

*mm*

^{2}, on the retina. The data to build the schematic eye was derived from the table reported in the original article.

^{21}Importantly, this approach aligns with the original methodology unlike that applied in previous studies.

^{18}

^{,}

^{19}Note the radius of the retinal sphere has been changed to match the one used for the histology map (

*r*= 11.459

_{0}*mm*, originally 11.06

*mm*in Drasdo and Fowler).

^{21}The distance between the center of the retinal sphere and the corneal vertex has also been scaled proportionally (

*c*= 12.38

_{0}*mm*, originally 11.95

*mm*). Therefore the default AL (AL

_{0}) of our schematic eye was 23.84

*mm*(originally 23.01

*mm*). These changes had a small impact on the degrees-to-

*mm*conversion, but a more important effect of the

*mm*

^{2}/solid degree ratio (Fig. 1). The schematic eye was coded in Matlab (The MathWorks, Natick, MA, USA). Additional details are reported in Appendix B.

*AL*, especially the size of the displacement area (see later) and the planar density of ganglion cells derived from histology. The assumption of a spherical shape for the retina for all ALs is prone to the adoption of a global expansion model. In this model, the planar RGC density would scale inversely with the square of the retinal radius, whereas the radius of the displacement zone (

*r*= 4.034

_{DZ}*mm*in the original paper)

^{15}would scale linearly with the retinal radius. The two scaling equations, where

*r*is the retinal radius corresponding to a given AL, are given as:

^{24}

^{–}

^{26}psychophysical evidence suggest that a global expansion model is a reasonable approximation for most axial lengths.

^{27}

^{–}

^{30}Global expansion is also implied in the RGC-RF model proposed by Drasdo et al.,

^{15}which assumes a constant density of RFs per solid visual degree (see later).

*mm*on the retina, should increase with axial length. This is a consequence of the stretching of the retinal tissue and Henle's fibers with increasing eye size. Although direct evidence of this is not yet available, indirect confirmation can be obtained by observing how the GCL profile scales with axial length in healthy eyes. To explore this, we used the 265 macular volume scans from Dataset 1 and identified the maximum GCL thickness peak for several meridians, centered on the anatomic fovea (Fig. 3). An ellipse was then fitted through a least squares method to the locations of the peaks. We then measured the length of the major and minor axes of the ellipse. All measurements were corrected for ocular magnification using the schematic eye defined in the previous section. The relationship between the length of the ellipse axes and the axial length was explored through linear regression. The ellipse dimensions were also predicted for an exact scaling with axial length, assuming a global expansion, by multiplying the ellipse dimensions predicted from the linear regression at

*AL*by the same scaling factor used for the

_{0}*r*(geometric scaling model). The goodness-of-fit of the linear regression and the geometric scaling model were compared using the mean absolute error (MAE), calculated for each model as the average of the absolute residuals.

_{DZ}*tpaps*function in Matlab) with anisotropic smoothing parameters, so that smoothing was stronger across B-scans than within a B-scan. The fovea was automatically identified through a template matching. Correct detection was confirmed through visual inspection.

^{12}to convert the OCT GCL thickness maps into customized estimates of RGC density and applied this to Dataset 2. In brief, the histology map was divided point-by-point by an average healthy GCL thickness map (768 × 768 pixels), obtained as the average from all eyes in Dataset 2, after aligning the fovea and the position of the optic nerve head (ONH). This yielded a volumetric density map (RGC/

*mm*

^{3}). The map can then be multiplied point-by-point by a GCL thickness map from a new subject to obtain a customized RGC density map (RGC/

*mm*

^{2}). We accounted for AL by applying a magnification correction to the GCL macular volume scans and by rescaling the histology density map according to the global expansion model given by Equation (2).

^{15}in their original article. The first step was to calculate the RGC-RF density along a specific meridian obtained from a generic model based on psychophysical measurements, the derivation of which is described in detail in the original article. The final formula, where

*e*is the eccentricity in visual degrees,

*D*is the density of RGC-RF (number/solid degree),

_{gcrf}*R*= 0.011785 and

_{v}*R*= 0.008333 and

_{o}*k*is a parameter that depends on eccentricity (see Appendix C), is given as follows:

*E*in Equation (3) was used by Drasdo et al.

_{2v}^{15}to scale the RGC-RF for each principal meridian. A key objective of our new approach was to determine its value for any arbitrary meridian. Similarly to Drasdo et al.,

^{15}we performed a numerical optimization of this parameter by simply requiring that the total counts of RGC-RF and RGC bodies are equal within the maximum displacement zone (

*DZ*). From Drasdo et al.,

^{15}the

*DZ*ends at 4.034

*mm*from the fovea and is assumed symmetric. This value was used for

*AL*and was scaled proportionally with the retinal radius for different axial lengths, as previously explained. The displacement is finally computed as the difference between the eccentricities at which the cumulative count of RGC bodies (

_{0}*C*) and the cumulative count of RGC-RF (

_{gcb}*C*) are equal. Additional details are reported in Appendix C.

_{gcrf}^{13}(Method 1) consisted of a simple displacement of the center of the stimuli, without any changes to its shape (Fig. 2, left panel). In the alternative method (Method 2), the circumference of the stimulus (approximated with 72 points around the stimulus edge) is displaced according to the Drasdo model; this results in distorted stimulus shapes in the parafoveal region (Fig. 2, right panel). We tested the accuracy of each method by requiring consistency under the Drasdo model. In fact, the model calculates the displacement by equating the number of expected RGC-RF and the number of RGC bodies at any given eccentricity (in a healthy eye). Therefore the estimated number of RGC-RF within a given stimulus area should match the number of RGC bodies within the displaced stimulus on the structural map, besides some minimal discrepancy due to approximation errors in the numerical calculations. For our calculations, we used a 10-2 grid and calculated the number of RGC-RF and cellular bodies for all conventional Goldman sizes, from I to V. The RGC-RF density function can change for each meridian. Hence, we generated a dense map with the same resolution as the structural map and used binary masks to calculate the number of RGC-RFs within each stimulus size. The same methodology was applied to the displaced stimuli on the structural map. The resolution used for our calculations was 2.2 µm (0.008°) for the histology map and 0.0391° for the structural maps in Dataset 2 (the maximum resolution of the Spectralis SD-OCT). This mainly affected the precision of the binary masks, which is important for small stimulus sizes. For the first analysis, we aimed for a very precise quantification to test the theoretical validity of the two methods. For the second analysis, we used a resolution that is likely to be applied for real data as a compromise between precision and speed of calculation.

^{31}for R

^{32}(R Foundation for Statistical Computing, Vienna, Austria). It allows the visualization of the schematic eye and the calculation for the RGC displacement. It can also import structural maps and provide calculations for different stimulus sizes. Finally, it can also be used for batch processing of a whole dataset. The App is freely available at https://relayer.online/drasdo, with detailed explanations on its use.

*ALs*(from 18 to 35

*mm*, at 0.5

*mm*intervals). The planar displacement maps were calculated out to 7.5

*mm*from the fovea, at 0.05

*mm*intervals, then organized in a dense three-dimensional array. The displacement values are then obtained via linear interpolation of the array.

*AL*before magnification correction (major axis MAE = 0.149

*mm*,

*P*< 0.001; minor axis MAE = 0.111

*mm*,

*P*< 0.001) and positively after magnification correction (major axis MAE = 0.148

*mm*,

*P*< 0.001; minor axis MAE = 0.112

*mm*,

*P*< 0.001). A geometric scaling model offered a very similar fit (major axis MAE = 0.175

*mm*; minor axis MAE = 0.131

*mm*). The results are presented in Figure 3.

*E*are reported in Figure 4A. Values are very similar to those reported by Drasdo et al.

_{2v}^{15}for the principal meridians for all considered

*ALs*. The systematic change with axial length was small (Fig. 4A). The displacement is constant for all axial lengths when measured in degrees, as a consequence of the global expansion mechanism assumed by the model. Figure 4 B shows the displacement in degrees and in

*mm*for

*AL*. The values in

_{0}*mm*are very similar to the average displacement reported by Drasdo et al.

^{15}

^{13}for healthy subjects. The same method was applied to the healthy macular volume SD-OCT scans in Dataset 2 (Fig. 6 and Table 2). The results were very similar to those obtained with the RGC histology map.

^{15}and studied its application to perimetric data to yield consistent structure-function measurements. We also developed a web App to make our methodology easily available for other researchers, in the hope to standardize this essential aspect of structure-function analyses.

^{19}

^{,}

^{20}we imposed weaker constraints on the symmetry of the displacement. The model proposed by Sjostrand et al.

^{20}used histological measurements to derive an even displacement around the fovea. Watson

^{19}followed an approach similar to Drasdo et al.

^{15}but used a different equation for the RGC-RFs and extended his calculations to arbitrary meridians by assuming an elliptical symmetry around the fovea. In contrast, our approach, as in the original article, only assumes the maximum displacement to be the same for all meridians in the fitting process. However, as shown in Figure 4B, such an assumption does not prevent the displacement from adapting to the measured distributions of RGC cells provided by histology. Importantly, the effective RGC displacement region extends to smaller eccentricities in the inferior retina. A similar approach for generalization of the Drasdo model has been proposed by Turpin et al.

^{18}Our results were in general agreement; they also showed a smaller extent of the displacement inferiorly compared to other regions. However, the displacement for the parafoveal locations was smaller in our calculations and in good agreement with the average displacement calculated by Drasdo et al.

^{15}In addition to previous work, we implemented a numerical ray tracing model of the schematic eye used by Drasdo and Fowler

^{21}to convert between visual degrees and distances on the retinal sphere. This allowed us to adapt the model so that the retinal sphere corresponded to the one used for the retinal histology map built by Curcio and Allen.

^{16}

^{,}

^{33}This is crucial to obtain consistent calculations, because the Drasdo model is based on that map. The implementation of the numerical model also allowed us to customize the conversion and the RGC density map based on the axial length. In this study, we assumed a global expansion model, scaling the linear structures with the radius of the retinal sphere and the density with the squared radius; this has been shown to be a good approximation by psychophysical examinations.

^{27}

^{–}

^{30}Additionally, we confirmed this by observing how the structure of the inner retina scales with axial length using a large dataset of SD-OCT data (Fig. 3). We found that geometric scaling for axial length fitted the observed data adjusted for ocular magnification. Under this assumption, the displacement is conveniently equivalent for all axial lengths when calculated in degrees of visual angle. However, competing models have been proposed for eye growth in myopia and an elliptical growth model, combining equatorial stretching and global expansion, seems to be the most realistic from anatomic studies.

^{24}

^{–}

^{26}One advantage of our numerical implementation of the schematic eye is that it can be easily adapted to accommodate for different types of expansion models. One major limitation of our structural dataset was the lack of extreme axial lengths. Determining the optimal expansion model with a stratified data collection of structural and functional data will be the objective of future work.

^{13}Unfortunately, those authors did not report tabulated RGC counts or estimated density. Nevertheless, the graphs reported in Figure 2 of their article

^{13}clearly show counts that, in healthy subjects, are compatible with the results of Method 1. For example, the largest RGC counts for a G-III stimulus were approximately 25.6 dB, very similar to our results in Figures 5 and 6 for Method 1 (25.5 dB). In turn, this was crucially less than half than that derived from Method 2 (29.5 dB) and the expected RGC-RF count from Equation (3) (29.4 dB). Moreover, the RGC-RF density derived from Method 1 for the smallest eccentricity (1.41°, Table 2), when substituted into Equation (3) to derive the corresponding visual acuity (with

*E*= 2), yields a value of 16.5 cycles/degree, unreasonably low for this eccentricity.

_{2v}^{15}In contrast, Method 2 yields 24.5 cycles/degree, much closer to the predicted 24.9 cycles/degree

^{15}and compatible with the literature.

^{34}These discrepancies might also be due to the fact that Yoshioka et al.

^{13}provided age-corrected structural and functional measurements at 64.5 years of age. However, the boxplots in the supplementary material for the same article show minimal changes between age-corrected and raw thickness values, too low to justify such a large difference.

^{14}

^{,}

^{23}Moreover, the actual methodology to implement the Drasdo model has been rarely reported. In many cases, the displacement appears to be symmetrical around the fovea.

^{7}

^{,}

^{13}

^{,}

^{22}

^{,}

^{35}

^{–}

^{38}This likely indicates an application of the average displacement profile presented in the graph from Figure 6 in the original paper by Drasdo et al.,

^{15}using a fixed degrees to

*mm*conversion. Although this might be satisfactory for some simple correlation analyses, disregarding the asymmetric nature of the displacement limits studies where a more detailed structure-function relationship is sought. For instance this is important when the development of a neural model of functional response is the main goal of the research.

^{13}In fact, as shown by our results in Figure 6, a high degree of consistency with the calculations can be achieved, especially considering that, like the Drasdo displacement model, the method to estimate the number of RGC cell bodies from structural measurements

^{12}is also based on the structural map produced by Curcio and Allen.

^{16}

**G. Montesano,**CenterVue (C);

**G. Ometto,**None;

**R.E. Hogg,**none;

**L.M. Rossetti,**CenterVue(C);

**D.F. Garway-Heath,**Carl Zeiss Meditec (C), CenterVue (C), Heidelberg Engineering (F), Moorfields MDT (P), ANSWERS (P), T4 (P);

**D.P. Crabb,**CenterVue (C), ANSWERS (P), T4 (P)

*Invest Ophthalmol Vis Sci*. 2012; 53: 3629–3637. [CrossRef] [PubMed]

*Graefes Arch Clin Exp Ophthalmol*. 2014; 252: 1621–1630. [CrossRef] [PubMed]

*Vis Res*. 2013; 82: 42–51. [CrossRef] [PubMed]

*Ophthalmology*. 2017; 124: 1449–1456. [CrossRef] [PubMed]

*Transl Vis Sci Technol*. 2016; 5: 15. [CrossRef] [PubMed]

*Progr Retinal Eye Res*. 2007; 26: 688–710. [CrossRef]

*Progr Retinal Eye Res*. 2013; 32: 1–21. [CrossRef]

*Invest Ophthalmol Vis Sci*. 2000; 41: 1774–1782. [PubMed]

*Invest Ophthalmol Vis Sci*. 2004; 45: 466–472. [CrossRef] [PubMed]

*Progr Retinal Eye Res*. 2010; 29: 249–271. [CrossRef]

*Perspectives in Glaucoma*. New Jersey, Thorofare, NJ: Slack, Inc.; 1988: 59–70.

*Invest Ophthalmol Vis Sci*. 2015; 56: 5548–5556. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2018; 59: 1693–1703. [CrossRef] [PubMed]

*Transl Vis Sci Technol*. 2018; 7: 36. [CrossRef] [PubMed]

*Vis Res*. 2007; 47: 2901–2911. [CrossRef] [PubMed]

*J Comp Neurol*. 1990; 300: 5–25. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2011; 52: 1486–1492. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2015; 56: 5984–5989. [CrossRef] [PubMed]

*Journal of Vision*. 2014; 14: 15. [CrossRef] [PubMed]

*Graefes Arch Clin Exp Ophthalmol*. 1999; 237: 1014–1023. [CrossRef] [PubMed]

*Br J Ophthalmol*. 1974; 58: 709–714. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2016; 57: 4815–4823. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2020; 61: 10. [CrossRef] [PubMed]

*Vis Res*. 2006; 46: 2236–2250. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2005; 46: 2698–2707. [CrossRef] [PubMed]

*Biomed Opt Express*. 2017; 8: 2386–2395. [CrossRef] [PubMed]

*Vis Res*. 2006; 46: 636–651. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2008; 49: 4679–4687. [CrossRef] [PubMed]

*Ind J Ophthalmol*. 2015; 63: 423–426. [CrossRef]

*Sci Rep*. 2020; 10: 12179. [CrossRef] [PubMed]

*R: A Language and Environment for Statistical Computing*. Vienna, Austria: R Foundation for Statistical Computing; 2019.

*Vis Res*. 1989; 29: 529–540. [CrossRef] [PubMed]

*J Vis*. 2011; 11: 13. [CrossRef] [PubMed]

*Am J Ophthalmol*. 2018; 194: 88–94. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2019; 60: 4241–4248. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2013; 54: 3046–3051. [CrossRef] [PubMed]

*Invest Ophthalmol Vis Sci*. 2014; 55: 5269–5277. [CrossRef] [PubMed]

*Optom Vis Sci*. 2011; 88: 4–7. [CrossRef] [PubMed]

^{33}the data were recorded by Curcio and Allen using spherical coordinates, reporting co-latitude (or retinal eccentricity) and longitude (or retinal meridian) in degrees. Eccentricity is calculated from the fovea, located at (0, 0). One degree of retinal eccentricity equals 2πr/360, where r is the radius of the assumed retinal sphere. From the header of the text file containing the tabulated data (https://research-materials.christineacurcio.com/), the assumed retinal sphere for the map is 11.459

*mm*. The map is built for a left eye, with the ONH at 20° eccentricity on the horizontal meridian (180° longitude). After conversion to

*mm*, a continuous histology map was obtained through cubic interpolation. Figure A1 shows the correspondence between the interpolated values and the original average counts (with standard deviations) provided by Curcio and Allen (https://research-materials.christineacurcio.com/) for the four principal meridians. The map was converted into a right eye by inverting the horizontal coordinates.

*ALs*, which can be achieved by proportionally changing the radius and the center location of the retinal sphere (Fig. 1). The anterior part of the schematic eye is left unchanged, so that the nodal point remains the same for all

*ALs*. The numerical estimate for the nodal point was obtained by averaging of 670 ray-tracings, with an angle of incidence between 0.1° and 67° with respect to the optic axis. The nodal point for each incident ray was derived numerically by minimizing the absolute difference between the angle of the incident and the refracted ray (after the posterior face of the lens) with respect to the optic axis. The resulting estimate was 6.93

*mm*, which is very close to the value of 6.95

*mm*reported by Drasdo and Fowler.

^{21}The schematic eye was assumed to be radially symmetric around the optic axis. The same schematic eye was used to correct for ocular magnification in the SD-OCT macular volume scans. In their original article, Drasdo and Fowler did not clarify what point they used as a reference to calculate the visual angle. In accordance with the guidelines from the Imaging and Perimetry Society,

^{39}we used the nodal point of the schematic eye.

*k*in formula 3 also depends on eccentricity according to

^{15}The

*D*is converted into a density per

_{gcrf}*mm*

^{2}with the conversion ratio derived from the schematic eye (Fig. 1), according to the axial length. The eccentricity in visual degrees is also converted into

*mm*on the retina using the schematic eye. Then, an RGC-RF density profile from the fovea outward can be calculated and converted into cumulative counts for a circular sector centered at the fovea by

*r*is the eccentricity from the fovea in

*mm*and

*2πr*is a correction factor for the sector area increasing with eccentricity. The actual width of the sector considered acts only as a scaling factor and has no bearing on the results of the computation. When computed numerically, the integral is simply a cumulative summation for predefined steps in

*r*. The same formula was used to compute the cumulative counts of RGC bodies (

*C*). In this case, the

_{gcb}*D*is simply replaced by the density of RGC bodies (per

_{gcrf}*mm*

^{2}) obtained from the histology map (

*D*) along the same meridian, after correction for the size of the retinal sphere. The displacement is finally computed as the difference between the eccentricities at which

_{gcb}*C*and

_{gcb}*C*are equal (Fig. A2). For each meridian, the displacement is zero beyond the first crossing point between the two cumulative curves. Therefore the actual displacement might involve a smaller area than the

_{gcrf}*DZ*.