Contrast Increment and Decrement Processing in Individuals With and Without Diabetes

Abstract

Purpose: Animal models suggest that ON retinal ganglion cells (RGCs) may be more vulnerable to diabetic insult than OFF cells. Using three psychophysical tasks to infer the function of ON and OFF RGCs, we hypothesized that functional responses to contrast increments will be preferentially affected in early diabetes mellitus (DM) compared to contrast decrement responses.

Methods: Fifty-two people with DM (type 1 or type 2) (mean age = 34.8 years, range = 18–60 years) and 48 age-matched controls (mean age = 35.4 years, range = 18–60 years) participated. Experiment 1 measured contrast sensitivity to increments and decrements at four visual field locations. Experiments 2 and 3 measured visual temporal processing using (i) a response time (RT) task, and (ii) a temporal order judgment task. Mean RT and accuracy were collected for experiment 2, whereas experiment 3 measured temporal thresholds.

Results: For experiment 1, the DM group showed reduced increment and decrement contrast sensitivity (F (1, 97) = 4.04, P = 0.047) especially for the central location. For experiment 2, those with DM demonstrated slower RT and lower response accuracies to increments and decrements (increments: U = 780, P = 0.01, decrements: U = 749, P = 0.005). For experiment 3, performance was similar between groups (F (1, 91) = 2.52, P = 0.137).

Conclusions: When assessed cross-sectionally, nonselective functional consequences of retinal neuron damage are present in early DM, particularly for foveal testing. Whether increment-decrement functional indices relate to diabetic retinopathy (DR) progression or poorer visual prognosis in DM requires further study.

Diabetic retinopathy (DR) is clinically diagnosed by identification of microvascular abnormalities within the inner retinal vasculature. There is, however, increasing interest in establishing novel clinical trial end points indicating visual function losses in DR (see review by Chen and Gardner, 2020),¹ given the accumulating evidence for diabetic retinal neuron damage possibly occurring prior to microvascular changes.^2,3 Common functional outcomes of diabetic retinal neurodegeneration include reduced contrast sensitivity,⁴ impaired color vision,⁵ reduced visual field sensitivity,⁶ and altered electrophysiological measures (i.e. multifocal electroretinogram and multifocal visually evoked potentials measures).^7–9 Some studies have also shown that functional deficits are correlated with structural thinning of the inner retinal tissues, indicating diabetic neurodegeneration.^10,11 

Diabetes-induced impairments in contrast sensitivity are attributed, in part, to retinal ganglion cell (RGC) damage. Recent work in experimental diabetic models suggest differential morphological and physiological alterations to ON- and OFF-type RGC: ON RGC in streptozocin mice showed a greater reduction in the dendritic field area and significantly altered passive membrane properties relative to OFF RGC.¹² In Ins2^Akita/+ diabetic mice, ON RGC were more likely than OFF RGC to suffer significant morphological changes when compared to control retinas, having 32.4% more dendritic terminals, 18.6% increase in total dendrite length, and 15.3% greater dendritic density.¹³ Although a specific mechanism for differential vulnerability of either ON or OFF RGC remains inconclusive, there is speculation from work in glaucoma models, that differences in vascular distribution between the ON and OFF sublaminae could be responsible for the susceptibility of one RGC type over the other.¹⁴ 

Visual perceptual detection of light and dark stimuli is thought to arise from ON and OFF parallel pathways with ON and OFF RGCs as key retinal constituents.^15–17 Hence, abnormalities in these RGCs may result in altered perception of contrast increments or decrements which in turn could be exploited as an index for diabetic damage to the retina. We sought to investigate whether diabetes mellitus (DM) differentially affected contrast increment or decrement responses using a battery of visual psychophysical paradigms designed to infer different features of the ON and OFF RGC functions. 

In healthy individuals, perception of lights and darks is asymmetric. Behavioral studies have confirmed that sensitivity to contrast decrements is greater than to contrast increments^18–20 and that decrement stimuli are detected faster than increment stimuli.^21,22 Likewise, cortical electrophysiological studies show a larger amplitude response for negative-contrast stimuli compared to positive-contrast stimuli, corroborating differential ON and OFF cortical contributions.^17,23 A decrement bias in perception is thought to relate a greater density of OFF RGCs compared with ON RGCs.^24–26 

We hypothesized that features of visual function related to contrast increments, as opposed to contrast decrements, would be more impaired in early DM. To test this, we used three psychophysical paradigms designed to measure different aspects of increment and decrement perceptual performance. Specifically, we assessed increment and decrement contrast sensitivity as well as visual temporal processing to increments and decrements using response time (RT) identification and temporal order judgment (which omits interindividual motor responses) in observers with and without DM. Our aim was to evaluate the potential for visual function indices targeting ON and OFF RGC as outcome measures for clinical trials, and to contribute to a more complete picture of the mechanisms behind diabetic-related visual impairment. 

Methods

Participants

Fifty-four participants with DM and 48 age-matched participants without DM were recruited. Participants were between the age of 18 and 60 years, had a Snellen visual acuity (VA) of at least 6/7.5 (at least 6/9 for participants with DM), with no more than +/−6.00 diopters (D) and +/−2.50 D of spherical and cylinder refractive error, respectively.^27,28 Individuals with DM had been diagnosed with either type 1 or type 2 DM and were using insulin and/or oral anti-hyperglycemic agents or had their diagnosis confirmed by one of the authors (S.F.). All participants that did not verbally report a recent 3-month glycated hemoglobin (HbA1c) were asked to undergo a finger-prick test using the Quo-Lab point-of-care HbA1c analyzer (EKF Diagnostics, Cardiff, United Kingdom); glucose levels were not measured at the time of testing. 

Exclusion criteria were as follows: the presence of co-existing ocular diseases that could affect the retina or vision, presence of diabetic macular edema (DME), having ever required retinal photocoagulation or administration of an intravitreal pharmaceutical agent to control DR or DME, pregnant or breastfeeding, people who experience migraines, use of psychotropic medication or any medication known to affect cognitive processing or vision, and individuals with a recent history of a cardiovascular event or on active treatment for cancer. DR severity was classified according to the International Clinical Diabetic Retinopathy and Diabetic Macular Edema Disease Severity Scales.²⁹ 

This study was approved by The University of Melbourne Human Research Ethics Committee (identifier: 1955267). All participants gave written informed consent. All procedures complied with the Declaration of Helsinki. 

For each participant, the study eye was the eye with the better VA; where VA was comparable between the eyes, the eye with the lower refractive error was studied. 

Equipment and Setup

Participants were seated at the appropriate distance for each component of the study from a gamma-corrected, Display++ LCD monitor (screen resolution = 1920 × 1080 pixels, dimensions = 39.5 cm [height] and 71 cm [width], and refresh rate = 120 hertz [Hz]) with their head supported on a chin rest to stabilize head movement. All experiments were conducted in a darkened room and the stimuli were presented with PsychoPy (version 3.2.3)³⁰ using custom software written in the Python interpreted language. Tasks were performed monocularly (the same fellow eye was occluded for the duration of each experiment) with optimal task-specific refractive correction. Participants were given time to practice the task until they and the examiner were confident that they understood the task. Practice sessions were predominantly 1 to 2 minutes in length as participants found the tasks easy to perform. 

Experiment 1: Contrast Sensitivity for Increments, Decrements, and Standard Gabors

Experiment 1 measured observers’ contrast thresholds to stimuli signaling: (a) standard contrast, (b) contrast increments, and (c) contrast decrements via a detection task. 

Gabor stimuli had a vertical orientation with a sinewave component modulated around: (a) mean luminance, (b) positive, or (c) negative phases of the sinusoid (Fig. 1A). Contrast for the mean luminance condition was defined as (Lmax-Lmin)/(Lmax-Lmin) whereas contrast increment and decrement Gabors were defined as the difference between the stimulus luminance (increment or decrement) divided by the background luminance (i.e. Weberian contrast). The spatial frequency of each Gabor stimulus was 3 cycles per degree (cpd), based on prior literature showing peak contrast sensitivity of 4 cpd at the fovea,³¹ and about 2 cpd in the retinal periphery.³² Each stimulus subtended 10 degrees × 10 degrees of visual angle and was presented in a spatially interleaved manner at the center of the screen and at 3 other eccentricities: 15 degrees nasal, 15 degrees above, and 15 degrees inferior to the central point of the screen. Participants were instructed to maintain fixation at a small (size = 5 pixels) black target while their stability of fixation was monitored via their reflection on a small mirror. 

Figure 1.

 

Contrast increment-decrement sensitivity task. (A) Vertically oriented Gabor stimuli with sinewave component modulated about the mean luminance, positive and negative phases of the sinusoid. Insert shows Gabors were presented with rapid onset then contrast ramped over 200 milliseconds. (B) Illustration of the experiment design where the stimuli were presented in an interleaved manner at four locations with 500 milliseconds (ms) wait time following either a “seen” or nonresponse.

Procedure

The time course of a single experimental trial is shown in Figure 1B. Gabor stimuli were onset then contrast ramped lasting 200 milliseconds (ms) in order to avoid triggering the opposite contrast.²⁰ Observers sat 50 cm from the screen, and were instructed to fixate at a small central target while responding via a keypad when they saw a Gabor stimulus. A maximum response window of 1.2 seconds (from the onset of the stimuli) was provided for the observer to make a “seen” response, which was then followed by 0.5 seconds wait time before the next stimulus presentation. Contrast thresholds were returned from a staircase procedure with a “1-up, 3-down” decision strategy where the staircase had 4 reversals and logarithmic unit step sizes of 0.2, 0.2, 0.1, and 0.1. Each run only contained one stimulus type. The contrast of the stimulus at the last 2 reversal intensities of each staircase were geometrically averaged and converted to contrast sensitivity in decibel (dB) units using the formula 10 × log₁₀(1/threshold). The final threshold was the average of the two runs per condition. 

Experiment 2: Visual Temporal Processing – Response Time Task

Experiment 2 measured observers’ response times to dark and light information. 

Dark or light suprathreshold targets subtended 0.5 degrees × 0.5 degrees of visual angle and were embedded in a random fashion on a centrally located 10 degrees × 10 degrees binary noise background. The design of this task was similar to the study by Komban et al. (2011).²¹ Suprathreshold targets were not individualized based on participant. Each pixel that made up the noise background subtended 0.08 degrees × 0.08 degrees at 1 m (Fig. 2A). 

Figure 2.

 

Response time (RT) task. (A) Illustration of the stimulus sequence showing light and dark targets on a noisy background and respective correct responses. (B)) RT for a single participant fitted to increment and decrement ex-Gaussian functions following averaging of RT histograms (bin size 0.2 seconds). An observer's mean RT to increments and decrements were then calculated as the sum of the exponential and Gaussian means, as described in Equation 1.

Procedure

Observers sat 1 m from the screen. One, 2, or 3 either dark or light targets were presented as in Figure 2A, whereas the observers were instructed to visually search and respond to the number of targets as rapidly as possible and as accurately as possible with a numerical keypad. Each presentation consisted of either light targets or dark targets. The duration of each presentation depended on the observer's RT. From pilot work, a total of 306 RTs (consisting of 153 dark and 153 light target presentations with equal numbers of 1, 2, and 3 target presentations for each contrast polarity) per observer was sufficient to show a consistent asymmetry between RT with dark and light targets. 

As shown in Figure 2B, RT histograms were averaged and fitted to an exponential-Gaussian (ex-Gauss) function in R.^33,34 Prior literature demonstrates that response times for behavioral tasks are typically well-represented by an ex-Gauss function, which is a convolution between an exponential (with a mean of τ) and Gaussian distributions (with a mean of µ and standard deviation of σ) defined by Equation 1.^34–36 The exponential component describes the time to decide on a response and the Gaussian distribution describes the time taken to perceive as well as initiate the physical response.³⁴ Hence, final response times are estimated as the combination of these two separate response times.³⁴ Ex-Gauss parameters µ, τ, and σ were estimated using the “optim” function in R³³ (which performs a minimization of the residual sum of squares), and the sum of µ and τ was calculated³⁴ as the estimate of the response time to be used in further analysis. Accuracy was calculated as a percentage of correct responses. 

\begin{eqnarray}f\left( t \right) &=& \frac{1}{\tau }\exp \nonumber \\ && \left\{- \frac{{\left( {t - {\rm{\;}}\mu } \right)}}{\tau } + {\rm{\;}}\frac{{{\sigma ^2}}}{{{\tau ^2}}}\right\} {\rm{*\;}}\oint \left\{ {{\rm{\;}}\frac{{t - {\rm{\;}}\mu }}{\sigma } - {\rm{\;}}\frac{\sigma }{\tau }{\rm{\;}}} \right\}\end{eqnarray}

(1)

 

Experiment 3: Visual Temporal Processing – Temporal Order Judgment Task

Experiment 3 measured thresholds for the ability of observers to judge differences in the temporal order of light and dark information. 

A pair of dark or light targets subtending 0.42 degrees × 0.42 degrees (20 pixels) of visual angle were presented on a 10 degrees × 10 degrees binary white noise background at a central location (Fig. 3A), in a similar design to Komban et al.²² Targets were presented either vertically or horizontally with reference to a red fixation point. The binary noise background was composed of pixels which subtended 0.02 degrees × 0.02 degrees. Dark and light pairs appeared with a variable offset in timing with a total of seven timing intervals (0–99.96 ms) presented a method of constant stimuli procedure. Temporal thresholds were taken as the 75% proportion of correct responses from the psychometric curve. 

Figure 3.

 

Temporal order judgment task. (A) Schematic illustrating the stimulus sequence where pairs of light and dark targets were presented with a variable offset in timing. The observer was required to judge which contrast target was presented first. (B) Psychometric curves for a single participant fitted to a cumulative Gaussian distribution, as described in Equation 2.

Figure 4.

 

Observers’ contrast sensitivity (dB) to increment, decrement and standard stimuli at central and three other eccentric locations (15 degrees from the center point) in the visual field. Means and confidence intervals for two groups: individuals with and without diabetes are represented. Different symbols highlight the types of diabetes: triangles for individuals without DM, filled diamonds for individuals with type 1 diabetes and open diamonds for individuals with type 2 diabetes.

Figure 5.

 

Observers’ response times (RT) to increment and decrement stimuli. Data for the two groups are shown as boxplots showing the median, 25th and 75th percentiles, minimum and maximum limits (denoted as points within 1.5 of the interquartile range from the edge of the boxplots). All individual data are shown; triangle, filled diamond, and open diamond symbols indicate participants without diabetes, participants with type 1 diabetes and participants with type 2 diabetes respectively. White and grey boxplot colors indicate increment and decrement targets.

Procedure

Observers sat 1 m from the screen and judged which contrast target (either dark or light) appeared first using a keypad (see Fig. 3A). Successive trials were initiated following the observer's response. Twenty presentations were provided for each timing interval per contrast, resulting in a total of 280 responses per observer. Psychometric functions were fit to the increment and decrement data separately for each individual using the Quickpsy package.³⁷ Functions were modeled with a cumulative Gaussian distribution (µ = mean, σ = standard deviation, and fp = false positive [defining the lower bounds of the curve, set as 0.5 for a 2AFC task as recommended by Wichmann and Hill],³⁸ fn = false negative or lapse rate [defining the upper bounds of the curve; Equation 2), where the proportion of correct responses were plotted against the timing intervals (Fig. 3B) 

\begin{eqnarray}\varphi \left( {x,{\rm{\;}}\mu } \right) &=& fp + \left( {1 - fp - fn} \right) \nonumber \\ && \times {\rm{\;}}\left[ {1 - GaussianCDF\left( {x,{\rm{\;}}\mu ,{\rm{\;\sigma }}} \right)} \right]\end{eqnarray}

(2)

 

Data Analysis

This study is part of a larger study, which is also evaluating structural imaging outcomes in DM. Thus, in designing the entire larger study, a power analysis was performed using G*Power (version 3.1.9.7)³⁹ based on the imaging data⁴⁰ (rather than functional data),⁴ as we expected our structural data to have an equivalent or smaller magnitude of effect relative to the differences in perceptual performance presented here. From structural optical coherence tomography – angiography data in Barraso et al. (2020),⁴⁰ for an effect size of 0.6, a proposed sample size of 60 observers per group would be required to achieve a power of 80%. A sample size based on functional performance (as in studies using contrast sensitivity),⁴¹ however, would require a smaller sample of 23 observers per group to achieve a similar power of 80%. 

IBM SPSS statistics software version 21.0 (https://www.ibm.com/analytics/spss-statistics-software) was used for statistical analysis. The Kolmogorov-Smirnov test was used to check whether the data met the assumption of normality. If the data were approximately normally distributed, we ran ANOVA tests to compare factors within and between our two groups. Mann-Whitney U tests were used where the data were not normally distributed. Correlation analyses were performed with Spearman rank correlation coefficient because some data did not meet normality assumptions. For all statistical testing, P < 0.05 was considered significant. 

Results

Fifty-four individuals with diabetes (mean age = 34.8 years) and 48 participants without diabetes (35.4 years) participated in this cross-sectional study. Demographic characteristics are shown in Table 1. 

Table 1.

 

Clinical Characteristics

 

Experiment 1: Contrast Sensitivity for Increments, Decrements, and Standard Gabors

Two individuals with DM were excluded from data analysis: one due to procedural error and the second reporting an inability to see the mean contrast Gabor stimuli even at the highest contrast, so no threshold was recorded. Of the remaining 52 individuals with diabetes, 32 had no DR, 12 had mild non-proliferative DR (NPDR), and 8 had moderate NPDR. 

In order to compare and confirm prior reports of contrast sensitivity differences foveally, we first compared contrast sensitivity at the central location for the standard stimulus.^41,42 This was statistically significantly different between individuals with DM (12.92 dB ± 2.32) and individuals without DM (13.98 dB ± 1.96), t (98) = 2.464, P = 0.015. However, when all 4 locations were included in the ANOVA, no difference was present between groups for the standard contrast sensitivity measure, F (1, 97) = 1.076, P = 0.302 (Figs. 4A–D). 

When contrast polarity (increment and decrement contrast sensitivity) and all 4 locations were considered, there was a between group effect where the group with DM had reduced contrast sensitivity, F (1, 97) = 4.04, P = 0.047. However, there was no interaction between groups and contrast polarity (P = 0.53) suggesting a reduction in both increment and decrement contrast sensitivity associated with DM. The interaction between location and group was significant, F (3, 291) = 3.25, P = 0.022, with the central location (Fig. 4A) revealing the largest contrast sensitivity group difference compared to the peripheral locations (Figs. 4B–C). Interindividual differences in performance were greater for mean contrast sensitivity compared to increment and decrement contrast sensitivity, especially in the peripheral locations. There was significant overlap between the groups, however, the worst performing people were almost exclusively from the DM group. Individuals with DM that fell outside the 95% limits of population control contrast sensitivity were numbered: increment = 5, decrement = 8, and mean = 7, and these individuals had a range of DR severity; 5 individuals with no DR, 1 with mild NPDR, and 2 with moderate NPDR. 

When considering age as a covariate, there was no main effect of age between groups (P = 0.195), and no significant interaction between age and location (P = 0.59) or contrast (P = 0.97). 

Experiment 2: Visual Temporal Processing – Response Time Task

Response accuracy for the DM group was 96.5% to dark targets and 89.9% to light targets; for individuals without DM, accuracy was 97.6% and 92.2% to dark and light targets, respectively. Only observers with a response accuracy above 80% were included in the RT portion of the analysis. One observer without DM and 6 observers from the DM group were excluded, leaving 47 participants without diabetes (mean age = 35.5 years) and 48 individuals with DM (mean age = 34.4 years). 

Mann-Whitney U tests were used as RT to increments and decrements within the groups were non-normally distributed. Both participant cohorts were significantly different (increments: U = 780, P = 0.01 and decrements: U = 749, P = 0.005) (Fig. 5). Overall, the control group had faster RT than the group with DM for both contrast polarities (group medians, increment: controls = 1.42 seconds, DM = 1.63 seconds and decrement: controls = 0.93 seconds, DM = 1.06 seconds). Moreover, outliers with slower response times tended to have type 2 diabetes. 

Experiment 3: Visual Temporal Processing – Temporal Order Judgment Task

The sample for this portion of the analysis was 50 observers with DM (mean age = 35.0 years) and 44 observers without DM (mean age = 34.8 years). Four observers from each group were excluded due to poor convergence of the curve fit the data, or evidence of inappropriate responses. These curve fits were (1) a poor match to the data (deviance³⁷: P < 0.05), (2) did not reach close to 100% at the longest time interval, and (3) produced thresholds that were less than 0 ms indicating that at 50% chance level, participants were heavily bias toward increments or decrements. 

Thresholds to increments and decrements were normally distributed for both groups. Visual inspection of Figure 6 shows a modest elevation of threshold in the DM group; however, a mixed ANOVA showed no main effect of group, F (1, 91) = 2.52, P = 0.137. Age as a covariate had a main effect, F (1, 91) = 13.80, P < 0.001 where older individuals had elevated decrement and increment thresholds. There was a significant main effect of contrast polarity indicating that thresholds to increment and decrement contrasts were different, F (1, 91) = 13.38, P < 0.001; however, there was no interaction between contrast polarity between the 2 groups (P = 0.451), suggesting that neither increment nor decrement visual temporal thresholds were preferentially reduced in DM. 

Figure 6.

 

Temporal thresholds to increment and decrement stimuli. Data are shown as means and 95% confidence intervals for each group, along with all individual data. Triangle and diamond symbols are represented by observers without diabetes and observers with diabetes, respectively.

Correlations Between All Three Tasks

We explored potential associations between different features of increment and decrement contrast processing by calculating correlation coefficients between our tasks in Table 2. The final sample for this analysis was 43 individuals without DM and 44 individuals with DM. 

Table 2.

 

Experiment Responses and Correlations

 

Significant correlations were found between standard contrast sensitivity with response times to decrements and with timing thresholds to increments. Likewise, a weak yet significant correlation was found between functional performance to decrements in experiments 2 and 3. 

Discussion

We hypothesized that diabetes-induced damage to the retina would affect contrast increment processing, reflecting preferential losses in ON RGC structure and function seen in murine models of DM. However, we found nonselective functional deterioration across two of three of our tasks, suggesting that in early DM, there are generalized losses in contrast increment and decrement functional performance. Key findings of our study support similar work identifying contrast information deficits as a proxy for retinal neuron damage in individuals with DM.⁴ Reduced contrast sensitivity (experiment 1) along with delayed RT to both contrast polarities (experiment 2) revealed diminished functional performance in individuals with DM. 

Gualtieri et al. (2011)⁴ also found nonselective functional deficits using psychophysical paradigms inferring distinct magno- and parvocellular pathways. Additionally, while Gualtieri et al. (2011)⁴ reported a lack of temporal processing deficits in their DM cohort, our results in experiment 2 revealed impairments to temporal processing in inferred ON and OFF RGC function. A lack of correlation between centrally measured contrast sensitivity, RT, and visual temporal thresholds suggests that the study paradigms used are measuring different aspects of contrast increment and decrement processing. To our knowledge, our work is the first to investigate correlations between tasks that measure increment and decrement processing in individuals with diabetes. Future work looking to evaluate inferred ON and OFF RGC function in DM should consider assessments that address both these contrast features as in experiments 1 and 2. 

In our sample of mainly young individuals with relative shorter duration of DM, foveal mean contrast sensitivity in experiment 1 showed a significant difference between observers with and without DM, and corroborates previous work showing reduced contrast sensitivity performance in DM.^10,41–44 Severity of mean contrast sensitivity deficits associated with DM can vary depending on how contrast sensitivity is measured as well as clinical characteristics of DM participants; factors such as age, type of DM,⁴¹ or duration of DM diagnosis⁴⁵ all contribute to the extent of contrast sensitivity loss. From our results, between group difference in mean contrast sensitivity only reached statistical significance when the superior location was omitted. Increment and decrement contrast sensitivity, instead of mean contrast sensitivity, might be a better functional measure in central and peripheral vision. A caveat here is that contrast sensitivity for both contrast polarities in the central location still had the largest group difference relative to peripheral locations, likely owing to a higher reliability of central vision.⁴⁶ 

Our sample was biased toward DM individuals at an early stage of DR (only a fraction [15%] of participants had moderate NPDR) which has advantages in the context of evaluating and developing novel approaches to detection of diabetic retinal damage or clinical trials end points for the vast population of individuals with DM who have no clinical signs of DR. Our cohort had good VA, more than half were classified as having no DR, and a large proportion had type 1 DM on mainstay insulin treatment. Emerging research shows that insulin has neuroprotective potential and treatment with insulin has been implicated to improve neuronal survival in DM⁴⁷ as well as promote recovery following neurotrauma or during neurodegenerative disease.⁴⁸ Agostinone et al. (2018)⁴⁹ also found that insulin treatment regenerated glutamatergic postsynaptic sites in ON and OFF RGC. However, our study was not designed to specifically assess differences in perceptual performance between individuals with type 1 or type 2 diabetes. Another potential confound was that we did not measure glucose levels at the time of testing, hence it is possible that some of the variability in performance in the diabetic group could be explained by variations in glucose levels.⁵⁰ 

We opted to use increment and decrement Gabor stimuli to measure the perceptual output of ON and OFF pathways (inspired by Luo-Li et al. 2016s²⁰ original experiments), reasoning that any damage to ON RGC would impact on the ON-signaling pathway and cortical responses to increments. These Gabor stimuli were specifically designed with a mean luminance background and were ramped across 200 ms to avoid stimulating the opposite contrast.²⁰ Although we cannot entirely be certain that increment and decrement Gabors isolate ON and OFF pathways, current understanding of the neurophysiology of human¹⁷ and non-human primate visual system^51–53 provides support for light and dark contrasts being separately processed by ON and OFF channels. A key observation in our data that supports differential pathway processing, at least in part, is that in experiment 1, measured thresholds differ for ON and OFF stimuli, replicating findings from Luo-Li et al. (2016).²⁰ 

Although all tasks showed an increment-decrement asymmetry, experiment 1’s results where increment contrast thresholds were lower than decrement contrast thresholds contradict evidence suggesting lower contrast thresholds to dark stimuli in comparison to light stimuli.^19,20,54,55 It is possible that our results may have been influenced by the type of experimental paradigm as experiments related to ON and OFF functional responses are (to our knowledge) only studied in central vision whereas experiment 1 involved testing contrast sensitivity in central and mid-peripheral visual field areas simultaneously. Because measured response differences between increments and decrements are typically small,²⁰ dividing the observer's attention outside the central vision could potentially have contributed to our different results. Shinomori et al. (2018)⁵⁶ used stimulus-onset asynchrony to derive impulse response functions for ON- and OFF-pathways and observed that the balance of sensitivities between increments and decrements also varied between participants. Meanwhile, stimuli parameters in experiments 2 and 3 were based on original experiments by Komban et al. (2011)²¹ and Komban et al. (2014),²² respectively, and were designed as centrally located, highly visible black and white targets embedded on a pixelated background composed of black and white targets. The results from these two experiments are consistent with previous findings in the literature^21,22 revealing that the visual system detects decrement contrast targets quicker than increment targets. 

In summary, we demonstrated that features of increment-decrement perceptual performance are impacted in early DM. Longitudinal work is required to assess the contribution of ON and OFF functional indices to the current diagnostic battery for diabetic retinal disease and to determine its prognostic value for improving visual outcomes for persons with DM. 

Acknowledgments

Disclosure: V.T.S. Tang, None; R.C.A. Symons, Novartis (C), Bayer (C), Mapkure (F), Ripple Therapeutics (F), Ionis Therapeutics (F), CSL (I); S. Fourlanos, None; D. Guest, None; A.M. McKendrick, None