Variability of the pattern electroretinogram

Documenta Ophthalmologica 70:103-115 (1988) �9 Kluwer Academic Publishers - Printed in the Netherlands

Variability of the pattern electroretinogram

KAREN HOLOPIGIAN, JEREMY SNOW, WILLIAM SEIPLE & IRWIN SIEGEL Department of Ophthalmology, NYU Medical Center, New York, N Y 10016, USA

Key words: amplitude variability, pattern electroretinogram, spatial frequency tuning

Abstract. Conflicting results have been obtained concerning the parametric properties of the pattern electroretinogram. These discrepancies may be due to the large amount of variability inherent in recording amplitudes. We have found the variability within a single stimulus .condition to be so large (ranging from 30% to 67% of the mean value) that it could mask any underlying spatial frequency tuning. Changing the stimulus conditions failed to significantly reduce the observed variability, although changing recording conditions produced some reduction. The use of a narrower rejection band, a greater number of sweeps, and placement of the reference electrode on the ipsilateral ear (as opposed to the ipsilateral temple) combined to decrease variability of the pattern electroretinogram within a single recording session; however, intersession variability remained high. Therefore one must be careful in evaluating data from this technique, and caution is advised in its clinical use.

Introduction

The pattern electroretinogram (PERG) reflects the electrical activity generated by the retina while an observer views a patterned stimulus. There is some experimental [1] and clinical evidence [2] that the activity measured by the PERG reflects the activity of the ganglion cell layer of the retina. If this potential originates from the proximal retina, then response amplitude should be sensitive to the distribtuion of pattern luminance, i.e., it should show spatial frequency tuning. In agreement with this, several studies have reported spatial frequency tuning of the PERG amplitude under a variety of stimulus conditions [3, 4, 5, 6]. Other investigators, however, have reported no spatial frequency tuning of the PERG, indicating that the PERG may entirely be a luminance response [7, 8, 9]. This discrepancy among the results of these experiments may stem from the large amount of amplitude variability observed in PERG recordings.

In this study we attempted to quantify the nature and sources of PERG amplitude variability. A replication of a previous study that reported PERG amplitude tuning was initially conducted [10]. We found that, even within

104

observers, the amplitudes of the positive and negative components of the PERG waveform were extremely variable, and any spatial frequency tuning, if present, was masked. Attempts to reduce this variability were made by changing various stimulus and recording conditions. The stimulus manipulations (e.g., reversal rate, rest periods, and pattern element size) did not appreciably reduce the amount of variability. Changes in the recording procedures (e.g., reference electrode placement, number of sweeps per tracing, reducing the sweep time, and increasing the signal-to-noise ratio), however, combined to decrease the variability within a session, although amplitude measures remained variable between sessions.

Methods

Three of the authors (KH, JS, WS) were observers in this experiment. All have normal visual acuity and were best corrected during the test.

The specific stimulus and recording parameters of this experiment were designed to replicate the study of Berninger and Schuurmans [10]. Square- wave grating patterns of varying spatial frequency (0.06, 0.25, 1, and 4 c/deg) were generated on a Conrac TV monitor (QQA 14/N) at a mean luminance of 50 cd/m 2. The background room luminance was 35 cd/m 2. The gratings (79% contrast) were counterphase reversed at either 4 or 10 reversals per sec (rps). The osbervers viewed the video monitor from a distance of 71 cm and the screen subtended 16 by 22 degrees.

Monocular PERGs were recorded with a gold foil electrode [11] referen- ced to the ipsilateral temple [12, 13]; the ipsilateral ear was grounded. Care was taken to ensure proper placement and stability of the electrode. The foil was positioned in the center of the lower lid and taped securely near the eye. The contralateral eye was patched. The raw signal was band pass-filtered between 0.3 and 30Hz and amplified (gain = 10,000) with a Grass P511J preamplifier. The amplified signal was averaged over a 500-msec epoch (N = 128) with artifact rejection (beyond __ 0.8 V at 1 V/div input sensitivity) on a Nicolet 1170B yielding two responses/sweep. The averaged responses were plotted and amplitudes were measured from baseline to the peak of the positive component (latency approx 50 msec) and from baseline to the trough of the negative component (latency approx 100 msec).

The sequence of presentation of spatial frequency was counterbalanced to avoid order effects. Only two replications were presented per day to minimize contamination by prolonged testing. In total, up to 12 replications per subject per condition were recorded.

Ampl i tude (uV) 2l

PERO POOtt2ve Component

O I I I O . O l O. I 0 1 . 0 0 I 0 . 0 0

SDOttal Frequency (c/deg)

105

Fig. 1. Average PERG amplitudes (+ 95% confidence intervals for the SEM) for each subject are plotted as a function of spatial frequency for the positive wave.

J~4~lltudo (uV)

o~ 0 , 0

PERO Negot lvo ColA~onont

I I I O . 1 0 1 . 0 0 I 0 . 0 0

~ a t l s l Froquoncy (c/deg}

Fig. 2. Average PERG amplitudes (_+ 95 % confidence intervals for the SEM) for each subject are plotted as a function of spatial frequency for the negative wave.

Ampl i tude (uV)

1 I 0 . 0 1 1 0 . 0 0

PERG P o s i t i v e COml3onent

O. :tO t . O0

106

Spe t l s1 Frequency (c/deg)

Fig. 3. The positive component amplitudes as a function of spatial frequency for six replications in one subject.

Results

Mean PERG amplitudes across sessions for each individual subject ( _ 95% confidence intervals) of the positive and negative components are plotted as a function of spatial frequency in Figs. 1 and 2. For the three subjects, mean amplitudes of the positive components did not show consistent change as a function of spatial frequency. The average amplitudes of the negative components showed spatial frequency tuning in two of the three subjects tested, with a peak at 1 c/deg and both a low and high spatial frequency fall-off. The third subject did not show tuning; instead, the mean amplitude decreased monotonically with increasing spatial frequency. When the PERG amplitude for the negative component was remeasured from the peak of the first positive wave to the following negative trough [14], these trends were identical. Although the negative component amplitudes appear to be tuned in the averaged data for two of three subjects, analyses of variance (ANOVAs) indicated that for any one subject there were no significant effects of spatial frequency for either the positive or the negative component.

Figs. 3 and 4 illustrate the variability among six replications for one subject (JS) for the positive and negative components, respectively. The amplitude changes as a function of spatial frequency were not consistent across replications for either component. The absent points in these figures

Amp l i t ude (UV)

0 t 0 . 0 1 10.00

PERG Neget tv8 Component

I I 0 . 1 0 I . O0

SpStlal Fpaquency (c/deg)

107

Fig. 4. The negative component amplitudes as a function of spatial frequency for six replications in one subject.

reflect not measurable data; such data points were not related to the order of testing or long-term electrode change, since subsequent recordings often yielded measurable responses.

The previous results emphasize the difficulty in conducting parametric studies on the PERG. We therefore attempted to identify the source(s) and quantify the magnitude of this variability by systematically examining its changes as a function of manipulations of the stimulus and recording conditions. Coefficients of variation (CV), that is, the standard deviation (SD) divided by the mean x 100 [15], were used to compare the magnitude of the variability among conditions where absolute amplitude might also vary. For example, the average coefficient of variation across replications, subjects and pattern element size for the data shown in Figs. 1 and 2 was 34% for the positive component and 40% for the negative.

The following sections describe various changes in the stimulus and recording parameters. In order to assess the effects of each experimental manipulation, a series of four replications for each subject was obtained within a single recording session and CVs were averaged across observers. As a baseline, CVs ranged from 30% to 67% among four replications using the stimulus (1 cpd) and electrode configurations described above. These changes are summarized in Table 1.

108

Table 1. P E R G manipulations.

P O S N E G

L Stimulus conditions 1. Acquisition. 10sec on, 10sec rest

2. Reversal rate. Increased to 10rps, steady-state

3. Effects of Blur. Half-screen alternating pattern

II. Recording conditions 1. E lec t rodes : E R G - j e t contact lens

Riggs scleral lens

2. Reference Electrode Position. Contralateral eye Ipsilateral ear

3. Signal-to-noise enhancement Signal acceptance interval reduced, N increased to 256

C V = 2 7 %

C V = 4 4 %

C V = 4 1 %

C V = 2 2 %

C V = 4 0 %

C V = 4 0 %

C V = 1 7 %

C V = 5 0 %

C V = 3 2 %

C V = 117%

C V = 2 6 %

C V = 3 0 %

C V = 4 0 %

C V = 2 3 %

C V = 1 8 %

Stimulus conditions

Fatigue. Stimulus presentations (1 c/deg, 79% contrast, 4 rps) were divided into 10-sec acquisition and 10-sec rest intervals which were repeated until the desired number of sweeps (128) was achieved. This was done in order to reduce subject fatigue, minimize fixation drifts, and lessen any variability which might be caused by ganglion cell adaptation, if present [16]. This did not reduce the variability in the amplitude of the positive component (CV = 27%) or the negative component (CV = 50%).

Steady-state stimulus. PERGs (1 c/deg, 79% contrast) were also measured in response to a faster reversal rate (10rps). The mean peak-to-peak amplitudes of these steady-state waveforms are large (2.5 to 4.8 #V), but the amount of PERG amplitude variability was comparable to the slower reversal rate, with an average CV of 32% across subjects.

Blur. Pattern blurring caused by accommodative and fixation changes during prolonged testing may also contribute to the observed variability. In order to eliminate the potential effects of blur, PERGs were recorded to a

109

O D

+2.5 D

+5.0 D

+10.0 D

0 D

Fig. 5. Averaged (N = 128) PERG waveforms of one subject obtained consecutively in a single session. The amount of stimulus blur was sequentially increased by the addition of positive diopter lenses. Note the amplitude variability both within and between conditions.

half-screen alternating pattern (79% contrast, 4 rps). In this condition the absolute amount of luminance change in the stimulus is the same as for the counterphase reversing gratings, but contour contributions are minimized (since there is only one border between the white and dark halves of the screen). The recorded response should be primarily derived from the local luminance changes, and the effects of blur should not affect this luminance potential. As expected, the average amplitude of the negative component was selectively reduced (although not extinguished) under this stimulus condition (average amplitude = 0.86~V); however, the CVs remained large (the average CV across subjects was 44% for the positive and 117% for the negative component).

The effects of blur were also examined more systematically. Here a 1 c/deg grating (4 rps, 79% contrast) was presented and the image was sequentially blurred with positive diopter lenses (from + 2.5 to + 10 D). If only the negative wave were pattern sensitive, as suggested by Berninger and Schuur- mans [10], increasing blur should decrease the amplitude of the negative wave, while that of the positive wave should be relatively unaffected. Two averaged responses were recorded per 500-msec tracing. Notice in Fig. 5 that the amplitudes of the first and second response of each tracing are not consistent for either the positive or the negative components and any trend

110

with increasing blur is obscured by this variability. Since each tracing was recorded under identical stimulus and electrode contact conditions, any difference between the first and second response on a single tracing is indicative of the variability in PERG recordings.

Since none of the above stimulus manipulations significantly reduced the observed variability of the PERG response, we next manipulated the electrode configuration and recording parameters in an attempt to examine the contribution of each on the variability of the PERG. The results of these attempts are described in the following sections and summarized in Table 1.

Recording conditions

Recording electrode. Both an ERG-jet (Life-Tech, Inc) contact lens electrode and a Riggs scleral lens electrode were used in an attempt to reduce variability that may be caused by corneal drying or unstable electrode contact due to displacement of the foil electrode by eye movements. Neither contact lens electrode, tested with a 1 c/deg, 79% contrast grating at 4 rps, increased the amplitude or decreased the amount of variability of the PERG. With the ERG-jet electrode the CV was 41% for the positive component and 26% for the negative, and with the Riggs electrode the CV was 22% for the positive component and 30% for the negative. These electrodes appear to have no advantage over the gold foil, whereas their discomfort and optical distortion are obvious disadvantages.

Reference electrode. The position of the reference electrode also was changed in order to examine the contribution of recording configuration on PERG amplitude and variability (1 c/deg gratings, 79% contrast, 4 rps).

The reference electrode was positioned in the contralateral eye (which was covered) as suggested by Maffei and Fiorentini [1] and Bodis-Wollner et al. [17]. This recording configuration has the advantage of reducing noise common to the two eyes. The average amplitude of both components was unchanged (average amplitude of 3.33 FtV for the positive and 2.52 #V for the negative component), but the variability was not reduced (CV = 40% for both components).

Simultaneous recordings were made to compare results obtained from the ipsilateral ear as a reference. Use of the ipsilateral ear increased the mean amplitude of the negative component and reduced the variability of the negative component to 23% (compared with 48% for the simultaneously recorded temple reference). The amplitude and CV for the positive com-

111

Amp l t t U d e (UV)

2

o 0.01

P o l i t t v i Colwonent

i i I O . l O 1 . 0 0 10 O0

S D S t l s l Fr lCluency ( c / d i g )

Fig. 6. Average PERG amplitudes (+ 95% confidence intervals for the SEM) for each subject obtained under our optimal recording conditions are plotted as a function of spatial frequency for the positive wave.

ponent were not significantly different between these two electrode configurations.

Signal-to-noise ratio. Because of the low amplitude of the PERG, any manipulation which increases the signal-to-noise ratio might be expected to reduce variability. Two recording parameters were changed in an attempt to increase the signal-to-noise ratio. Artifact rejection was increased by reducing the acceptance band from -t-0.8 volts to + 0.4volts, and the number of sweeps per average was increased to 256. Sweep time also was decreased to 150 msec to reduce the amount of direct current drift in the records. These manipulations decreased within session variability to an average of 17 % for the positive component (average amplitude = 1.92#V) and 18% for the negative component (average amplitude = 1.62 #V).

On the basis of the above findings we repeated the original spatial frequency tuning experiment using stimulus and recording conditions which had minimized PERG variability within a replication series. The ear was used as the reference, sweep time was 150 msec, a greater number of responses were averaged for each condition (N = 256), and the acceptance interval was decreased. Six replications per subject were collected over three sessions. For these data the mean amplitude of the positive component increased as pattern element size increased (Fig. 6). For only one subject, the mean

112

Ae~ l I tu~e (uV)

PEFG Neglttve Coeponent

I O. ~0

0 I I 0 . 0 1 t . 0 0 I0.00

S p a t i s l Frequency (c/deg)

Fig. 7. Average PERG amplitudes (+ 95% confidence intervals for the SEM) for each subject obtained under our optimal recording conditions are plotted as a function of spatial frequency for the negative wave.

amplitude of the negative component indicated spatial frequency tuning (peaking at 1 c/deg). The other subjects showed no tuning of the mean negative component amplitudes (Fig. 7).

While the changes noted above reduced variability within a recording session, intersession variability remained high, with the average CV across subjects and conditions 26% for the positive component and 43% for the negative component. We have not identified the source of this residual variability.

Discussion

It is clear from the results of the present experiments that our attempts to minimize PERG amplitude variability were not successful enough to allow parametric examination of the response properties. We were able to reduce the amount of amplitude variation within a single recording session; however, intersession variability remained high. Because of inconsistencies in P E R G amplitudes, any trends that might be present in the d/tta were obscured.

Possible causes of this large amount of variability in PERG amplitudes include inconsistencies in stimulus parameters and recording techniques.

113

1 cpd

4 cpd PERG

1 cpd

4 cpd NASION

1 cpd

4 cpd O C C I P U T

Fig. 8. Simultaneous recordings of the PERG and VEP (measured at the nasion and occiput). When the spatial frequency was changed from 4 to 0.5 c/deg, the phase of the VEP recorded at the occiput shifted about 180 deg, whereas the phase of the responses recorded at the nasion and the PERG remained in phase, showing no shift with spatial frequency.

Stimulation factors such as subject fatigue, fixation drifts, possible ganglion cell adaptation, and blurring were examined, but none of our manipulations reduced the variability of the PERG response. It is noted that simultaneously recorded visual evoked potential (VEP) and P E R G differ greatly in the degree of their variability. The VEP has an average CV of 15%, whereas the PERG recorded at the same time has an average CV of 36% for both the positive and negative components.

Amplitude variability may also arise from recording problems such as unstable electrode contact (e.g. arising from electrode movement or corneal drying), reference configuration, and signal-to-noise ratios. We found that increasing the number of sweeps, decreasing the sweep time rate, and placement of the reference electrode on the ipsilateral ear of the subject produced less variable recordings within a single recording session; however, the intersession variability remained high.

The use of the ipsilateral ear as a reference for the P E R G has been criticized for presumed contamination of the recorded waveform caused by the cortical evoked potential [13]. These authors have demonstrated that at one temporal frequency (5 rps) the negative component of the PERG ap- pears to contain VEP components from the nasion, since the waveforms recorded from the eye and the forehead are in phase. We have found, however, that the VEP does not contaminate the PERG. Although under certain stimulus conditions it is possible to obtain VEP and PERG responses which are in phase, when temporal or spatial characteristics of the stimulus

114

are changed a phase shift of the VEP as recorded at the occiput results; however, the phase of the PERG and the response recorded at the nasion do not show a corresponding phase change (Fig. 8). Additionally it has been documented that the spatial frequency and temporal tuning of the PERG differ from that of the VEP [3, 16, 18] - further evidence that the VEP does not significantly contaminate the PERG. The potential recorded at the nasion reflects retinal activity rather than VEP activity; see Adachi-Usami et al. [19] and Seiple and Siegel [20].

Previous PERG studies have indicated large deviations in amplitude measurement, although none have subjected their data to statistical analysis. Schuurmans and Berninger [12] present data with an average CV of 20% to 35%. Korth and Ilschner [14] show data with CVs averaging 37% for pattern onset PERG responses and 55% for pattern offset response. Korth and Rix [21] showed data which have CVs of 31% for the positive component and 26% for the negative component. Hess et al. [22] present contrast thresholds and PERG amplitudes in a group of 34 normal observers as a function of spatial frequency. The averaged CVs for their psychophysical estimates of contrast thresholds were 20%, while the averaged CVs for their PERG amplitudes were larger than 100%. Trick and Trick [23], although reporting good intrasubject replications of the shapes of the tuning functions, present data that show large amplitude variations across sessions, with CVs in the 30% range. Finally, Mehaffey (personal communication) has found CVs ranging from 15% to 50% in PERGs recorded from anesthetized monkeys.

Because of this great variability, one must be careful to evaluate PERG recordings, and caution is advised in the clinical use of the PERG.

References

1. Maffei L, Fiorentini A. Electroretinographic responses to alternating gratings before and after section of the optic nerve. Science 1981; 211: 953.

2. Arden GB, Vaegan, Hogg CR. Clinical and experimental evidence that pattern electroretinogram (PERG) is generated in more proximal retinal layers than the foveal electroretinogram (FERG). Ann NY Acad Sci 1982; 388: 580.

3. Odom JV, Maida TM, Dawson WW. Pattern evoked retinal response (PERR) in human: Effects of spatial frequency, temporal frequency, luminance, and defocus. Curr Eye Res 1982; 2: 99.

4. Hess RF, Baker CL. Human pattern-evoked electroretinogram. J Neurophysiol 1984; 51: 939.

5. Arden GB, Vaegan. Electroretinograms evoked in man by local uniform or patterned stimulation. J Physiol 1983; 341: 84.

6. Sokol S, Jones K, Nadler D. Comparison ofthe spatial properties ofthe buman retina and cortex as measured by simultaneously recorded pattern ERGs and VEPs. Vision Res 1983; 23: 723.

115

7. Armington JC, Brigell M. Effects of stimulus location and pattern upon the visually evoked cortical potential and the electroretinogram. Int J Neurosci 1981; 14: 169.

8. Reimslag FCC, Ringo JL, Spekrejise H, Verduyn LH. The distinction between luminance and spatial contrast components in the pattern ERG. Doc Ophthalmol Proc Series 1983; 37: 255.

9. Kirkham TH, Coupland SG. Pattern ERGs and check size: absence of spatial frequency tuning. Curr Eye Res 1982/83; 2: 511.

10. Berninger T, Schuurmans RP. Spatial tuning of the pattern ERG across temporal frequency. Doc Ophthalmol 1985; 61: 17.

11. Arden GB, Carter RM, Hogg C, Siegel IM, Margolis S. A gold foil Electrode: Extending the horizons for clinical electroretinography. Invest Ophthalomol Vis Sci 1979; 18: 421.

12. Schuurmans RP, Berninger T. Luminance and contrast responses recorded in man and cat. Doc Ophthalmol 1985; 59: 187.

13. Yanashima K, Yoshi M, Okisaka S. The relation between the after-negative potential of the pattern electroretinogram and the visually evoked cortical potential. Doc Ophthalmol 1986; 63: 137.

14. Korth M, Ilschner S. The spatial organization of retinal receptive fields in light and darkness as revealed by the pattern electroretinogram. Doc Ophthalmol 1986; 63: 143.

15. Sokal RR, Rohlf FJ. Biometry. (W.H. Freeman, 1969): San Francisco. 16. Odom JV, Norcia Am. Retinal and cortical potentials: spatial and temporal characteris-

tics. Doc Ophthalmol Proc Series 1984; 40: 29. 17. Bodis-Wollner I, Bobak P, Harnois C, Mylin L. Methodological aspects of signal-to-noise

evaluation of simultaneous pattern electroretinogram and visual evoked potential recordings. Doc Ophthalmol Proc Ser 1983; 40: 39.

18. Peachey NS, Seiple WH. Contrast sensitivity of the human pattern electroretinogram. Invest Ophthalmol Vis Sci 1987; 28: 151.

19. Adachi-Usami E, Kuroda N, Nakajima I. Distribution of pattern-evoked potentials in the facial area. Am J Ophthalmol 1983; 96:734.

20. Seiple W, Siegel I. Recording the pattern electroretinogram: a cautionary note. Invest Ophthalmol Vis Sci 1983; 24: 796.

21. Korth M, Rix R. Changes in spatial selectivity of pattern-ERG components with stimulus contrast. Graefe's Arch Clin Exp Ophthalmol 1985; 223: 23.

22. Hess RF, Baker CL, Verhoeve JN, Keesey UT, France TD. The pattern evoked electroretinogram: Its variability in normals and its relationship to amblyopia. Invest Op- hthalmol Vis Sci 1985; 26: 1610.

23. Miller RF, Dowling JE. Intracellular responses of the M/iller (Glial) Cells of the mud- puppy retina: their relation to b-wave of the electroretinogram. J Neurophysiol 1970; 33: 323-41.

Address for offprints: K. Holopigian, Dept. of Ophthalmology, 550 First Avenue, NYU Medical Center, New York, NY 10016, USA.

Variability of the pattern electroretinogram

Documents

Transcript of Variability of the pattern electroretinogram