Abstract
Purpose:
The purpose of this study was to characterize summation of temporal L- and M-cone contrasts in the parvo- (P-) and magnocellular (M-) pathways in glaucoma and the relationship between the respective temporal contrast sensitivities (tCS) and clinical parameters.
Methods:
Perifoveal tCS to isolated or combined L- and M-cone contrasts (with different contrast ratios, and therefore different luminance and chromatic components) were measured at different temporal frequencies (at 1 or 2 Hz and at 20 Hz) using triple silent substitution in 73 subjects (13 healthy, 25 with glaucoma, and 35 with perimetric glaucoma). A vector summation model was used to analyze whether perception was driven by the P-pathway, the M-pathway, or both. Using this model, L- and M-cone input strengths (AL, AM) and phase differences between L- and M-cone inputs were estimated.
Results:
Perception was always mediated by the P-pathway at low frequencies, as indicated by a median phase angle of 179.84 degrees (cone opponency) and a median AL/AM ratio of 1.04 (balanced L- and M-cone input strengths). In contrast, perception was exclusively mediated by the M-pathway at higher frequencies (input strength not balanced: AL/AM = 2.94, median phase angles = 130.17 degrees). Differences in phase were not significant between diagnosis groups (Kruskal-Wallis = 0.092 for P- and 0.35 for M-pathway). We found differences between groups only for the M-pathway (L-cone tCS deviations at 20 Hz were significantly lower in the patients with glaucoma P = 0.014, with a strong tendency in M-cones P = 0.049). L-cone driven tCS deviations at 20 Hz were linearly correlated with perimetric mean defect (MD) and quadratically correlated with retinal nerve fiber layer (RNFL) thickness.
Conclusions:
Unaltered phase angles between L- and M-cone inputs in glaucoma indicated intact temporal processing. Only in the M-pathway, contrast sensitivity deviations were closely related to diagnosis group, MD, and RNFL thickness, indicating M-pathway involvement.
Glaucoma is a heterogeneous group of chronic-neurodegenerative diseases of the retinal ganglion cells and among the most common causes of blindness worldwide.
1,2 The current therapeutic approach of decreasing intraocular pressure
3,4 is often not sufficient, and the need for neuroprotective strategies is widely recognized.
5
For the development and evaluation of such therapies, new tests of ganglion cell-related visual function are required that are more sensitive than the current gold standard,
6 which is standardized automated perimetry (SAP) of the central 30 degrees using a white Goldman III
4 stimulus on a white background.
7 Improved diagnostic performance has been achieved with several psychophysical methods that are supposed to isolate certain subpopulations of retinal ganglion cells (and their respective retino-geniculate pathways).
8,9
The two dominant subpopulations of retinal ganglion cells in the human retina are the midget and the parasol ganglion cells, which differ in morphology, connectivity, receptive fields, spatial distribution, and physiological properties.
10,11
The midget ganglion cells project to the parvocellular layer of the lateral geniculate nucleus (LGN; P-pathway) and are therefore also called P ganglion cells (P-GCs). They receive inputs from midget-bipolar cells and mediate red-green color vision by opponent L- and M-cone inputs.
11 Although most humans have more L- than M-cones, their inputs to the midget ganglion cells are balanced, probably due to synaptic weightings. Thus, sensitivities to L- and M-cone isolating stimuli are equal when perception is mediated by the P-pathway.
12 The P-pathway has a high spatial but low temporal resolution.
The parasol ganglion cells project to the magnocellular layers of the LGN (M-pathway, M-GCs).
11 They transfer additive signals from L- and M-cones, resulting in a strong luminance sensitivity, and their input strengths are proportional to their packing densities. The M-pathway mediates luminance perception with a high temporal but low spatial resolution.
13
The axons and cell soma of the MC ganglion cells might be more susceptible to glaucomatous damage because they are larger.
14 Although some studies found psychophysical evidence for such preferential damage,
15,16 other studies did not.
17–20 For the P-pathway, it seems that studies that use red-green contrast for isolation instead of high spatial / low temporal resolution tasks for isolation even found larger functional defects in the P-pathway.
21–23 Furthermore, it is not known whether any of these techniques might be able to demonstrate functional consequences of reduced synaptic density before ganglion cell death.
24,25
We hypothesized that characterizing the processing of photoreceptor inputs in the retinal ganglion cells helps to clarify the mechanisms behind such conflicting results.
The properties of the P- and M-pathways can be assessed by using stimuli with different L- and M-cone contrast ratios.
26 Large luminance contrasts (to which particularly the M-pathway is sensitive) can be created by in-phase stimulation of the L- and the M-cones, large chromatic contrasts (to which the P-pathway is sensitive) by stimulation in counterphase. Most stimuli (including L- and M-cone isolating stimuli) will have a luminance and a chromatic component. It has been shown before that flicker detection thresholds for simultaneous stimulation of L- and M-cones can be predicted from the thresholds to single cone isolating stimuli based on a vector summation model.
27,28 The L- and M-cone isolating stimuli can be created using the silent substitution technique. The method has been described in detail before
29–31 and will be reviewed briefly in the Methods section.
It was the purpose of the present study to investigate (1) if detection thresholds of different combinations of L- and M-cone stimulation can be explained by the same vector summation model in patients with glaucoma and in healthy subjects, (2) how pathologic changes in glaucoma affect model parameters, and (3) how glaucomatous alterations in temporal contrast sensitivities relate to changes in retinal nerve fiber layer (RNFL) thickness and to SAP field sensitivity losses. In an earlier study, Alvarez et al.
32 used a very similar model for analyzing red-green contrasts, but they have not explicitly considered vector summation of the L- and M-cone inputs as the underlying mechanism and they have not validated their assumption that the major and minor axis of the fitted ellipses reflect activities of the P- and M-pathways.
Therefore, we examined flicker detection thresholds in an annular perifoveal test field for sinusoidal modulation of L- and M-cone isolating stimuli and for well-defined combinations of these in three groups: healthy subjects, glaucoma suspects without manifest visual field defects, and patients with perimetric glaucoma.
Altogether, data from 73 subjects were included in this study. These data were collected in two series of measurements. The main series of measurements – using mixed L- and M-cone- stimuli with different L:M ratios (including L- and M-cone isolation) at 2 Hz and 20 Hz – was made in the year 2017, and data from these subjects were used in all analyses in this study (
n = 57, 13 healthy subjects, 25 glaucoma suspects, and 19 patients with perimetric glaucoma; see
Table 1). Healthy subjects were recruited among the staff of the University Hospital Erlangen and among healthy relatives of the participating patients with glaucoma.
Table 1. Demographic and Clinical Data of the 2017 Cohort
Table 1. Demographic and Clinical Data of the 2017 Cohort
In addition, data from another cohort of patients with glaucoma (
Table 2) were used together with data from the first cohort for the comparison between contrast sensitivities and the clinical parameters perimetry and ocular coherence tomography (OCT) RNFL thickness. These data had already been obtained in 2015 (
n = 20, 4 subjects had participated in both studies). All patients in this second cohort had perimetric glaucoma and visual field defects in the central 12 degrees, and only temporal contrast sensitivities to pure L- or M-cone-isolating stimuli at 1 Hz and 20 Hz were measured. Preliminary data from both series were shown at the annual ARVO conventions in 2016 and 2019.
33,34
Table 2. Demographic Data From the Additional Perimetric Patients With Glaucoma From the 2015 Cohort
Table 2. Demographic Data From the Additional Perimetric Patients With Glaucoma From the 2015 Cohort
Table 3. Parameters Used for Data Analysis and How They Were Calculated
Table 3. Parameters Used for Data Analysis and How They Were Calculated
This study followed the tenets of the declaration of Helsinki and was approved by the Ethics committee of the medical faculty of the Friedrich-Alexander - University Erlangen - Nürnberg. All subjects gave written informed consent.
All patients with glaucoma participated in an ongoing longitudinal, observational study of glaucoma, the Erlangen Glaucoma Registry. Patients were examined annually under standardized conditions. Therapeutic decisions were made by the treating physicians independent of participation in the present study. Best-corrected visual acuity was measured with Snellen charts at a distance of 5 m and converted into logMAR. Further routine examinations included Haag-Streit slit lamp examination, indirect funduscopy with a 78-diopter-lens, and gonioscopy with a Goldman-three-mirror-lens.
Patients with congenital (X-chromosomal) color vision defects were excluded from the study. We considered a normal test Farnsworth Panel D15 result sufficient (saturated version). An anomaloscope examination was always performed if there were any errors in the D15 (Rayleigh equation, HMC anomaloscope; Oculus, Wetzlar, Germany) and also in the majority of subjects with a normal D15. Exclusion criteria were (1) presence of other retinal diseases, especially age-related macular degeneration, (2) diabetes mellitus independent of the presence of diabetic vasculopathy, (3) use of medications that impair visual function, and (4) clinically relevant cataract. An experienced ophthalmologist examined all subjects that were included in this study. Patients with diabetes mellitus were excluded because neural damage may precede diabetic vasculopathy. Normal age-related yellowing of the lens should not impair L- and M-cone isolating temporal contrast sensitivities according to calculations that we have previously published.
35
Subjects were divided into three groups: (1) healthy subjects, (2) glaucoma suspects, and (3) perimetric glaucoma patients. Healthy subjects were characterized by an IOP of ≤ 21 mm Hg, normal Octopus G1 visual field (mean deviation ≤ −2 dB and absence of relevant focal defects, defined as 3 adjacent fields with a corrected probability of functional loss of ≥ 5%), and a normal morphology of the optic disc. Glaucoma suspects had normal visual fields, but had either ocular hypertension (IOD ≥ 21 mm Hg) or characteristic glaucomatous changes of the optic disc (cupping, focal defects of the neuroretinal rim, or the nerve fiber layer), or both. Patients with perimetric glaucoma had correlating optic disc changes and visual field defects. Differences in the visual field global indices (
P < 0.001) and RNFL thicknesses (
P < 0.001) between groups were a direct consequence of the group definitions, and therefore of no relevance. LogMAR was not significantly different between the groups. There were statistically significant age differences between groups (see
Table 1;
P < 0.001), which were addressed by correcting contrast sensitivities for age (see below).
Briefly, the photoreceptor types differ in their spectral sensitivity, that quantifies the probability that a photon of a given wavelength is absorbed. However, the principle of univariance
29 states that the reaction of the photoreceptor is always the same independent of the photon's wavelength. Thus, when two stimuli of different spectral compositions are exchanged, differences in a photoreceptor's sensitivity to the stimuli can be compensated by choosing appropriate intensities. For instance, an exchange between two stimuli, one containing short wavelengths and the other long wavelengths, may not result in a modulation of L-cone excitation when the short wavelength stimulus is more intense than the long wavelength stimulus. However, the same stimulus will strongly stimulate the M-cones. In that case, the response of any system that exclusively or mainly receives L- and M-cone input (as the M- and P- pathways at photopic conditions) will be mediated by the M-cones. When four primaries with different wavelength contents are used, perfect isolation of each of the four photoreceptor types can be achieved by varying intensity of all four primaries (triple silent substitution) at identical states of retinal adaptation.
29,30
We previously validated that photoreceptor isolation is feasible.
12,35,39 As mentioned above, L- and M-cone isolating stimuli contain both a luminance and a chromatic component and detection thresholds were mediated by the P-pathway at low and by the M-pathway at high temporal frequencies.
12,41
Using matrix calculation, modulation contrasts and phases (either in-phase or counter-phase) of the LEDs were calculated based on the spectral sensitivities of the photoreceptors in a way that L- and M-cones were stimulated at different L:M contrast ratios (again, either in-phase or counter-phase), whereas contrasts in S-cones and rods were either zero (silent substitution) or at least negligible (because of variability e.g. in pigment spectra or preretinal absorption
35). The LED contrasts and phases that we used are shown in the
Supplementary Table S1, together with the resulting contrasts at the photoreceptor level.
Photoreceptor-isolating temporal modulation thresholds were measured at a convenient time during a visit for the standardized routine examinations. Measurements were carried out in a separate, calm, and dimly lit room. All subjects were adapted to the room light for at least 15 minutes prior to measurements. We measured the more severely affected eye unless there were exclusion criteria, and we chose the right eye when there was no relevant difference between eyes. The fellow eye was occluded with a transparent eyepatch and the subjects were positioned in front of the LED stimulator. Breaks were taken every 15 minutes.
We have described psychophysical threshold determination in detail before.
12 In brief, we used a PEST strategy with two randomly interleaved staircases. One staircase started at 0%, the other at 100% contrast.
In the 2015 cohort, only thresholds to pure L- and M-cone isolating stimuli were measured (L:M ratios of 1:0 and 0:1). In the 2017 cohort, at least four out of eight ratios were measured (1:0, 0:1, 1:1, 1:-1, 1:2, 2:1, 1:-2, and 2:-1; negative ratios indicate counterphase modulation).
If the contrast thresholds for mixed L- and M-cone stimuli can be described by vector summation of the L- and M-cone inputs with a phase lag, the measured thresholds are located on an ellipsis around the origin in a plot where L-cone contrast C
L is plotted against M-cone contrast C
M (
Fig. 2).
14,24,25
Our vector summation model is based on the following equations, which calculate C
L and C
M for a given L:M ratio (
\({\rm{K}} = \frac{{{{\rm{C}}_{\rm{L}}}}}{{{{\rm{C}}_{\rm{M}}}}}\)). For pure L-cone isolating stimuli, C
M = 0 and
\({{\rm{C}}_{\rm{L}}} = \frac{1}{{{{\rm{A}}_{\rm{L}}}}}\). For all other ratios, the following equations are used:
\begin{equation}{{\rm{C}}_{\rm{M}}} = \frac{1}{{\sqrt {{\rm{A}}_{\rm{L}}^2 \cdot {{\rm{K}}^2} + 2 \cdot {{\rm{A}}_{\rm{L}}} \cdot {{\rm{A}}_{\rm{M}}} \cdot {\rm{K}} \cdot {\rm{cos\Delta \alpha }} + {\rm{A}}_{\rm{M}}^2} }}\end{equation}
and
\begin{equation}{{\rm{C}}_{\rm{L}}} = {\rm{K}} \cdot {{\rm{C}}_{\rm{M}}}.\end{equation}
The model parameters are AL and AM, which represent the sensitivities of the two photoreceptor types, and Δα, which is the phase lag between both. For pure L- or M-cone-isolating stimuli, contrast sensitivities (CS) were defined as the inverse of the threshold contrast for that photoreceptor type. Thus, according to the equations above, the contrast sensitivities CSL and CSM for these stimuli would equal the model parameters AL and AM for a perfect model fit.
Nonlinear least-squares estimate for the model parameters were obtained using the nlsLM function of the R-package minpack.lm. The underlying nonlinear function returned the cartesian distance of a threshold from the origin (\(\sqrt {C_L^2 + C_M^2} \)) as a function of K based on the equations shown above.
We have excluded measurements where observers were not able to perceive maximally possible stimulus contrasts, because this would have resulted in floor effects of the resultant sensitivities. If less than four points were available for analysis, a model fit was not performed (NA values).
Peripapillary RNFL thickness was measured using spectral domain ocular OCT with the HRA system (Heidelberg Engineering, Heidelberg, Germany). We averaged the RNFL thickness from sectors that contribute to the test field by averaging the RNFL thickness of 7 out of 32 sectors (RNFL6 degrees). These were sectors 2, 3, 4, 5, 28, 30, and 31, where sector 1 is located just above the temporal horizontal hemi-meridian and sectors are numbered clockwise.
In 3 healthy subjects and one glaucoma suspect from the 2017 cohort, contrast thresholds measurements were performed for 8 different L:M ratios (see
Fig. 3). At both 2 Hz and 20 Hz, we were able to describe contrast sensitivity by the above-mentioned model based on vector summation.
At 2 Hz, contrast thresholds were generally lower (and sensitivities were thus higher) than at 20 Hz. This corresponds to a position closer to the origin in
Figure 3. For all 4 subjects, the fits to the 2 Hz data revealed that Δα was close to 180 degrees at 2 Hz and that A
L and A
M had similar values (thus, the ratios
\(\frac{{{{\rm{A}}_{\rm{L}}}}}{{{{\rm{A}}_{\rm{M}}}}}\) were close to 1). Altogether, the model fits were excellent at 2 Hz (residual sum of squares = median 0.07, and range = [0.0008 to 25.4]).
In contrast, at 20 Hz, as mentioned above, values for A
L and A
M were generally smaller than at 2 Hz, there were marked interindividual variabilities in the phase angles, and model fits were poorer (residual sum of squares: median 0.25, range = [0.006 to 13.3], Wilcoxon
P = 0.038 compared with 2 Hz). In all 4 cases shown in
Figure 3, the estimated L-cone-driven sensitivity (A
L) was higher than that driven by the M-cones (A
M) and thus the ratios
\(\frac{{{{\rm{A}}_{\rm{L}}}}}{{{{\rm{A}}_{\rm{M}}}}}\) were larger than 1. Compared with the three healthy subjects, the glaucoma subject had much higher thresholds and thus smaller A
L and A
M values at 2 and 20 Hz.
For the rest of the 2017 cohort, we determined contrast thresholds for four different L:M stimulation ratios (0:1, 1:0, 1:1, and 1:-1). The results of the model fits are shown in
Figure 4. Despite the limited number of data points, model fits were generally satisfactory. At 2 Hz, the estimates for the parameters A
L and A
M were, again, very similar to each other for each subject (median L:M ratio = 1.04), and the phase angles were close to 180 degrees (median phase angle = 179. 84; see
Fig. 4, top panels), regardless of diagnosis group (Kruskal-Wallis:
P = 0.092). This confirms mediation of temporal contrast perception by the parvocellular system. Indeed, the subjects reported substantial chromatic changes in the test field close to threshold.
At 20 Hz, the subjects were more sensitive for L-cone than for M-cone stimuli in most cases (median L:M ratio = 2.94). Phase angles Δα were much more variable but generally substantially smaller than 180 degrees (median = 130.17 degrees; see
Fig. 4, bottom panels). Again, this was the case for all groups. This is in agreement with the notion that thresholds were mediated by the magnocellularly based luminance channel and by the fact that the subjects reported that at threshold they perceived achromatic flicker. Although the absolute sensitivities were lower at high temporal frequencies compared to low temporal frequencies, the subjects generally reported to be more certain about their answers.
There were no significant differences in phase angle between diagnosis groups (Kruskal-Wallis test: P = 0.092 for 2 Hz and P = 0.35 for 20 Hz). Furthermore, there were excellent correlations between the model parameters AL and AM on the one hand and the observed contrast sensitivities for pure L- and M-cone isolating stimuli (CSL and CSM) on the other hand. This confirmed the theoretical relationship from the above equations and highlighted the quality of the model fits (Spearman Rho: 0.90 for L-cones 2 Hz, 0.97 for L-cones 20 Hz, 0.95 for M-cones 2 Hz, and 0.98 for M-cones 20 Hz, P < 0.001 in all cases). Therefore, we decided to use the L- and M-cone contrast sensitivities (converted to sensitivity deviations in dB) because this allowed us to include data from patients with perimetric glaucoma measured in 2015.
In 2015, the low frequency measurements were performed at 1 Hz, but according to previous studies, sensitivities at 1 Hz and at 2 Hz are quite similar.
12 We used a model for age correction that we developed in an earlier study.
40
The sensitivity deviations for the different diagnosis groups are shown in
Figure 5. At high temporal frequencies, there was a statistically significant loss of temporal contrast sensitivity in the perimetric glaucoma group.
As mentioned before, the losses were age corrected. We also recalculated the expected differences assuming a larger age-related loss of 0.15 dB/year instead of 0.1 dB/year, because confidence intervals for this slope indicated a 5% probability of an age effect of 0.15 or larger, but the differences between groups remained significant (Kruskal-Wallis P = 0.041 instead of 0.014 after correction for multiple testing).
At low temporal frequencies, photoreceptor-isolating sensitivity deviation values were quite variable, and there was considerable overlap between the different diagnosis groups. There was a tendency toward a sensitivity loss in the glaucoma suspects with four glaucoma suspects showing substantial sensitivity losses, despite normal visual acuity, visual field, and anomaloscope findings (Rayleigh equation).
Our data show that retinal processing in glaucoma patients can be characterized using combined L- and M-cone-isolating stimuli. The model, which assumes vector addition of the photoreceptor signals in the retinal ganglion cells, can indeed describe flicker detection thresholds for such stimuli. The data indicate that all stimuli – regardless whether chromatic or luminance contrast was dominant – were detected by the P-pathway at low and by the M-pathway at high temporal frequencies. We did not find evidence for relevant changes in retinal processing in glaucoma. Altogether, this implies that the sensitivities to temporal L- and M-cone contrasts can be used as parameters for functional changes in these systems. Among these parameters, temporal contrast sensitivity for L-cone isolating stimuli at 20 Hz (M-pathway) was significantly reduced in patients with advanced glaucoma. Furthermore, exclusively this sensitivity was significantly correlated with perimetric MD6 degrees (linear relationship) and RNFL6 degrees (quadratic relationship).
Supported by German Research Council (DFG) Grants HU2340/1-1 to CH and KR1317/16-1 to JK. Research Grants from the Friedrich-Alexander-University Erlangen-Nürnberg (ELAN 11.03.15.1, IZKF Rotationsstelle).
Disclosure: C. Huchzermeyer, None; F. Horn, None; R. Lämmer, None; C. Mardin, None; J. Kremers, None