Diabetic retinopathy (DR) is the most frequent cause of vision loss among working-age adults in the Western world.1 DR has classically been considered a microcirculatory disease of the retina and the clinical signs of DR primarily involve structural changes of the retinal vasculature, including microaneurysms, hemorrhages, and vascular occlusions. These vascular abnormalities can result in functional losses, particularly in cases of edema, capillary non-perfusion, and retinal detachment. Visual acuity (VA), which is the most common measure of visual function in patients with DR, can be reduced when the vascular abnormalities affect the central macula.
In addition to the well-documented vascular changes, neural and optical abnormalities have also been described in patients with DR. Apoptosis has been reported in the retinal ganglion cell layer in these patients2,3 and the loss of cells results in a reduction of the retinal nerve fiber layer thickness.4,5 Furthermore, neural function may be indirectly affected by abnormalities in retinal glia that have been reported in animal models of DR including reactive gliosis,6 impaired neurotransmitter uptake,7 and altered metabolism.8 In addition to neural dysfunction, diabetic patients can have optical defects that include structural changes of the cornea and lens9 and elevated higher-order optical aberrations.10
The neural abnormalities can alter retinal signal processing and optical abnormalities can degrade the retinal image by reducing contrast and introducing phase shifts. These neural and optical abnormalities may lead to an elevated level of blur within the visual system, which is expected to adversely impact visual function. However, the relative contribution of optical and neural blur to VA loss in patients with DR is not well understood at present. One approach to evaluate the relative contributions of these sources of blur is to measure the total amount of blur within the visual system (i.e., “equivalent intrinsic blur”)11,12 and then partition the equivalent intrinsic blur measurement into optical and neural components.13 An estimate of optical blur can be derived from the optical point spread function (PSF), whereas neural blur can be defined as the blur that remains once the contributions of optical blur to equivalent intrinsic blur are accounted for. Thus, this definition of neural blur includes neural factors such as apoptosis, as well as factors that indirectly affect neural function such as glial abnormalities, ischemia, microaneurysms, and edema.
The purpose of the present study was to characterize the relationships among VA, equivalent intrinsic blur, optical blur, and neural blur in patients with proliferative diabetic retinopathy (PDR). Measurements obtained from patients with PDR were compared to a group of normally sighted individuals to determine the extent to which equivalent intrinsic blur, optical blur, and neural blur differ from normal.
The study was approved by an institutional review board at the University of Illinois at Chicago and conformed to the tenets of the Declaration of Helsinki. Written informed consent was obtained from each subject. Ten subjects diagnosed with PDR (seven males and three females, ages 25 to 68 years) and 10 normally sighted control subjects (five males and five females, ages 46 to 63 years) participated in the study. The mean age (± standard deviation, SD) of the subjects with PDR (48.5 ± 12.6 years) was not significantly different from that of the controls (54.4 ± 4.9 years) (t = 1.38, p = 0.18). Subjects with PDR were recruited from the Retina Service at the University of Illinois at Chicago. The subjects had a diagnosis of clinically significant diabetic macular edema (CSDME), vitreous hemorrhage (VH), and/or central retinal vein occlusion (CRVO). Table 1 provides the characteristics of the subjects, including sex, age, diagnosis, and chart acuity as assessed with the Lighthouse Distance Acuity Chart.
The pupil of the tested eye was dilated with 2.5% phenylephrine hydrochloride drops prior to testing. All measurements were made after determining the best optical correction of low-order aberrations (second order) and the subjects viewed the stimuli through a 3-mm artificial pupil to control the retinal illuminance. Control subjects had chart VA that ranged from −0.16 to 0.06 log MAR (approximately 20/14 to 20/23 Snellen equivalent) and the subjects with PDR had chart VA that ranged from 0.00 to 0.66 log MAR (approximately 20/20 to 20/90 Snellen equivalent).
The methodology for the psychophysical measurement of equivalent intrinsic blur (σ int) is described in detail elsewhere.13,14 In brief, test stimuli consisted of unblurred or blurred tumbling E optotypes. The blurred Es were generated by convolution with 2D low-pass Gaussian functions of four different standard deviations (σ stim): 0.0 (unblurred), 0.4, 1.6, and 6.5 arcmin. Threshold log MAR for each value of σ stim was determined using a two-alternative forced-choice staircase procedure. Threshold log MAR values were plotted as a function of log σ stim and were fit with the log form of the following equation12:
where MAR0 represents VA for the unblurred target (σ stim = 0) estimated by the fit and σ int represents equivalent intrinsic blur. MAR0 and σ int were free parameters that were adjusted to minimize the mean squared error between the data and the fit. Equivalent intrinsic blur was defined as the value of log σ stim that increased log MAR0 by 0.15 log units (i.e., log √2, which is the knee-point of the fit), consistent with a previous definition.12,13
Optical blur was derived from Shack-Hartmann wavefront measurements as described previously.13 Briefly, the wavefront aberration function for high order (third- to sixth-order) aberrations was measured over a 3-mm pupil diameter. From these measurements, the two-dimensional optical PSF was derived according to standard transformations,15 radially averaged to provide a one-dimensional line profile, normalized to unity, and fit with a Gaussian function. The standard deviation of the best-fit Gaussian function defined optical blur (σ opt). The Gaussian function provided an excellent description of the PSF over the central region from the peak to 1.5 arcmin (mean ± SD R2 value of 0.99 ± 0.002; N = 20).
The neural blur index (η) was defined as:
The value of η provides an index of blur that is introduced by neural (i.e., non-optical) sources. When η is equal to 0.50, the optical and neural contributions to σ int are equal, whereas higher values of η are associated with increased neural contributions to σ int.
Log MAR Versus Log σ stim
Fig. 1 presents log MAR VA as a function of log σ stim for the 10 subjects with PDR (the symbols representing the individual patients correspond to those given in Table 1) compared to the range of normal (gray region). The corresponding Snellen equivalents of the log MAR values are shown on the right y axis for reference. The range of normal was derived by first determining the maximum and minimum threshold log MAR for each σ stim for the 10 control subjects and these data were then fit with Eq. 1. The curves fit to the patients’ data are the least-squares best fits of Eq. 1. This function transitions from a slope of 0 at low values of log σ stim to a slope of 1 at high values. Equation 1 provided an excellent fit to the data for the subjects with PDR and for the control subjects, yielding a mean ± SD R2 value of 0.96 ± 0.05 (N = 20).
For subjects with PDR, threshold log MAR for the unblurred E (MAR0; σ stim = 0) spanned a broad range (0.82 log units). Despite the variation in log MAR0 for the subjects with PDR, the mean log MAR0 ± SD for the PDR subjects (0.42 ± 0.26) was significantly greater (t = 5.38, p < 0.05) than that of the control subjects (−0.05 ± 0.09). As σ stim was increased, the log MAR differences among the PDR subjects decreased, such that the data points for individual patients tended to converge for the most blurred E (MAR6.5; σ stim = 6.5 arcmin). Thus, log MAR6.5 spanned a relatively narrow range (0.17 log units) and the mean log MAR6.5 for the subjects with PDR was significantly greater than that of the control subjects (t = 9.71, p < 0.05). Of note, there were three subjects with PDR (numbers 1–3) who had normal or nearly normal log MAR0 (log MAR0 of 0.14 or less), but log MAR6.5 that was outside of the normal range.
Equivalent Intrinsic Blur
Fig. 2 shows the relationship between log MAR0 and log σ int for the subjects with PDR and the controls. Data for the control subjects are represented by the open circles and data for the subjects with PDR are represented by the symbols corresponding to those given in Table 1. As in Fig. 1, the corresponding Snellen equivalents of the log MAR values are shown on the right y axis for reference. Log MAR0 increased linearly with log σ int (slope = 1.32) and there was a significant correlation between the two parameters (r = 0.93, p < 0.05; N = 20). The three PDR subjects (numbers 1–3) with relatively good log MAR0 all had normal σ int. The remaining seven PDR subjects had substantial log MAR0 elevations (log MAR0 of 0.38 or greater) and all had elevated levels of σ int. The mean ± SD σ int for the 10 subjects with PDR (2.03 ± 0.98 arcmin) was significantly higher (t = 2.88, p < 0.05) than that of the control subjects (1.12 ± 0.22 arcmin).
Fig. 3 shows the relationship between log MAR0 and log σ int for the subjects with PDR and the controls. There was no statistically significant correlation between log MAR0 and log σ opt for the combined patient and control data (r = 0.32, p = 0.17; N = 20) or for correlations performed separately on the control data (r = 0.04, p = 0.92; N = 10) and patients’ data (r = 0.48, p = 0.16; N = 10). Furthermore, the mean σ opt for the PDR subjects did not differ significantly from that of the control subjects (0.40 ± 0.02 and 0.40 ± 0.01 arcmin for the patients and controls, respectively; t = 0.69, p = 0.50).
Fig. 4 shows the relationship between η and σ int. Data for the control subjects are represented by the open circles and data for the subjects with PDR are represented by the symbols corresponding to those given in Table 1. The dashed horizontal line at η = 0.50 represents equal contributions of optical and neural blur to σ int. The solid curve is a prediction derived from Eq. 2 by varying σ int between 0.60 and 4.2 arcmin and setting σ opt equal to 0.40 arcmin (the mean σ opt for the PDR and control subjects). Because the value of σ opt was constant in the fit, increasing σ int results in an increase in the relative contribution of neural blur to σ int. As expected, this curve provided an excellent description of the data (R2 = 0.99; N = 20), substantiating that as σ int increased, the relative contribution of neural blur to σ int increased for the PDR and control groups. The data points of nearly all subjects fell above 0.50, indicating that σ int had a greater contribution from neural blur than optical blur. The values of η for the control subjects were typically between 0.50 and 0.75. In comparison, 7 of 10 subjects with PDR (the same seven subjects with elevated log MAR0; Figs. 1 and 2) had values above 0.75, indicating a large contribution of neural blur to σ int. In contrast, the three subjects with PDR who had normal or near-normal log MAR0 also had η within the range of normal. Mean η of the PDR subjects was significantly greater than that of the controls (0.63 ± 0.08 and 0.75 ± 0.14 for the patients and controls, respectively; t = 2.23, p = 0.04).
Evidence from human subjects with DR and from animal models of DR suggests that elevations in both neural and optical blur could contribute to visual dysfunction. Therefore, it is of interest to evaluate equivalent intrinsic blur, optical blur, and neural blur in these patients to determine the relative contributions of these sources of blur to VA reductions. Subjects with PDR who had substantial VA reduction (log MAR0 of 0.38 or greater) had elevated equivalent intrinsic blur levels compared to visually normal control subjects. Furthermore, our index of neural blur (η) was also abnormally high in these patients, suggesting that elevated neural blur is an important factor in their VA loss. In contrast, these patients all had values of optical blur that were within the range of normal.
The similar levels of optical blur caused by higher-order aberrations in the PDR and control subjects suggests that optical blur had a negligible contribution to the VA loss of the PDR subjects in the current study. A previous report10 showed small but significant elevations in optical blur in diabetic patients compared to visually normal control subjects, but that study did not evaluate VA. The apparent discrepancy between the two studies may be attributed to the PSF calculation over a larger pupil size (6 mm), as compared to the current study (3 mm), or to differences in patient samples.
Although all of the patients had PDR, their clinical manifestations varied, with patients having CRVO, VH, and/or CSDME. Despite these differences, log MAR0 was highly correlated with log σ int in our sample of patients. The 4 patients with ETDRS chart VA better than approximately 20/25 had CSDME that did not affect the fovea, which likely accounts for why these 4 patients had relatively well-preserved VA. Three of these 4 patients (numbers 1–3) also had normal or near-normal log MAR0, σ int, σ opt, and η. Four other patients (numbers 5–8) with VH had moderate losses of ETDRS chart VA (20/33 to 20/48) and elevated σ int and η. The 2 patients with the worst ETDRS chart VA of our sample had CSDME that affected the fovea (number 9) and CRVO (number 10), respectively, and these 2 patients had elevated σ int and η. The source of the elevated η in patients with VA loss is presently uncertain, as several factors could elevate η, including apoptosis, glial abnormalities, edema, and/or ischemia. Future work is needed to define the relative contributions of these factors to the VA loss in patients with PDR.
There were two notable findings for VA measurements made with the most blurred E (log MAR6.5) for the patients with PDR. First, there was relatively small variation in log MAR6.5 among these subjects, which is in contrast to their substantial variation in log MAR0. A likely explanation for this finding is that log MAR6.5 was not dependent on equivalent intrinsic blur because the amount of stimulus blur (6.5 arcmin) exceeded the amount of blur within the visual system. Once the effects of equivalent intrinsic blur were overcome by the high stimulus blur, the VA values for the subjects with PDR were similar. There was also less variation in log MAR6.5 (0.19 log units) than in log MAR0 (0.28 log units) for the control subjects.
The second finding of interest for VA made with the most blurred E is that log MAR6.5 was outside of the normal range for all subjects with PDR, including those who had normal or near-normal log MAR0 (subjects 1–3). Thus, from a practical view, measurement of VA for highly blurred letter optotypes may be useful for some patients with PDR because it can reveal VA deficits that are not apparent by conventional chart acuity testing. Future studies are needed to identify the source of the VA deficit for the highly blurred optotypes, but the deficit is unlikely related to optical or neural blur because the amount of stimulus blur exceeded these levels. Additional work is also needed to investigate the clinical utility of measuring VA for highly blurred optotypes to determine if this measure of visual function could be applied as an outcome measure in clinical trials or for monitoring disease progression in patients with PDR.
J. Jason McAnany
Department of Ophthalmology and Visual Sciences
University of Illinois at Chicago
1855 West Taylor St
Chicago, IL 60612
This research was supported by National Institute of Health grants R00EY019510 (J.J.M.), R01EY014275 (M.S.), P30EY001792 (departmental core grant), the Department of Veterans Affairs (M.S.), the Cless Family Foundation Fund for Retina Research (J.L.), and Research to Prevent Blindness unrestricted departmental and Senior Scientific Investigator (M.S.) awards.
Received: July 1, 2013; accepted September 3, 2013.
1. Fong DS, Aiello L, Gardner TW, King GL, Blankenship G, Cavallerano JD, Ferris FL 3rd, Klein R. Diabetic retinopathy. Diabetes Care 2003; 26: 226–9.
2. Abu El-Asrar AM, Dralands L, Missotten L, Geboes K. Expression of antiapoptotic and proapoptotic molecules in diabetic retinas. Eye (Lond) 2007; 21: 238–45.
3. Barber AJ, Gardner TW, Abcouwer SF. The significance of vascular and neural apoptosis to the pathology of diabetic retinopathy. Invest Ophthalmol Vis Sci 2011; 52: 1156–63.
4. Lopes de Faria JM, Russ H, Costa VP. Retinal nerve fibre layer loss in patients with type 1 diabetes mellitus without retinopathy. Br J Ophthalmol 2002; 86: 725–8.
5. van Dijk HW, Verbraak FD, Kok PH, Garvin MK, Sonka M, Lee K, Devries JH, Michels RP, van Velthoven ME, Schlingemann RO, Abramoff MD. Decreased retinal ganglion cell layer thickness in patients with type 1 diabetes. Invest Ophthalmol Vis Sci 2010; 51: 3660–5.
6. Rungger-Brandle E, Dosso AA, Leuenberger PM. Glial reactivity, an early feature of diabetic retinopathy. Invest Ophthalmol Vis Sci 2000; 41: 1971–80.
7. Puro DG. Diabetes-induced dysfunction of retinal Muller cells. Trans Am Ophthalmol Soc 2002; 100: 339–52.
8. Lieth E, Barber AJ, Xu B, Dice C, Ratz MJ, Tanase D, Strother JM Glial reactivity and impaired glutamate metabolism in short-term experimental diabetic retinopathy. Penn State Retina Research Group. Diabetes 1998; 47: 815–20.
9. Ishiko S, Yoshida A, Mori F, Abiko T, Kitaya N, Konno S, Kato Y. Corneal and lens autofluorescence in young insulin-dependent diabetic patients. Ophthalmologica 1998; 212: 301–5.
10. Shahidi M, Blair NP, Mori M, Zelkha R. Optical section retinal imaging and wavefront sensing in diabetes. Optom Vis Sci 2004; 81: 778–84.
11. Watt RJ, Morgan MJ. Spatial filters and the localization of luminance changes in human vision. Vision Res 1984; 24: 1387–97.
12. Levi DM, Klein SA. Equivalent intrinsic blur in spatial vision. Vision Res 1990; 30: 1971–93.
13. McAnany JJ, Shahidi M, Applegate RA, Zelkha R, Alexander KR. Contributions of optical and non-optical blur to variation in visual acuity. Optom Vis Sci 2011; 88: 716–23.
14. McAnany JJ, Alexander KR, Lim JI, Shahidi M. Object frequency characteristics of visual acuity. Invest Ophthalmol Vis Sci 2011; 52: 9534–8.
15. Thibos LN, Hong X, Bradley A, Applegate RA. Accuracy and precision of objective refraction from wavefront aberrations. J Vis 2004; 4: 329–51.