Abstract
purpose. The interindividual variability of the visual evoked potential (VEP)
has been recognized as a problem for interpretation of clinical
results. This study examines whether VEP variability can be reduced by
scaling responses according to underlying electroencephalogram (EEG)
activity.
methods. A multifocal objective perimeter provided different random
check patterns to each of 58 points extending out to 32° nasally. A
multichannel VEP was recorded (bipolar occipital cross electrodes, 7
min/eye). One hundred normal subjects (age 58.9 ± 10.7 years)
were tested. The amplitude and inter-eye asymmetry coefficient for each
point of the field was calculated. VEP signals were then normalized
according to underlying EEG activity recorded using Fourier transform
to quantify EEG levels. High α-rhythm and electrocardiogram
contamination were removed before scaling.
results. High intersubject variability was present in the multifocal VEP, with
amplitude in women on average 33% larger than in men. The variability
for all left eyes was 42.2% ± 3.9%, for right eyes 41.7% ± 4.4%
(coefficient of variability [CV]). There was a strong correlation
between EEG activity and the amplitude of the VEP (left eye, r = 0.83; P < 0.001; right
eye, r = 0.82; P < 0.001).
When this was used to normalize VEP results, the CVs dropped to 24.6%±
3.1% (P < 0.0001) and 24.0% ± 3.2%
(P < 0.0001), respectively. The gender difference
was effectively removed.
conclusions. Using underlying EEG amplitudes to normalize an individual’s VEP
substantially reduces intersubject variability, including differences
between men and women. This renders the use of a normal database more
reliable when applying the multifocal VEP in the clinical detection of
visual field changes.
Recent advances in clinical
electrophysiology
1 2 3 4 5 6 7 8 9 10 have extended the possible
applications of the visual evoked potential (VEP) to objective
visual-field mapping. However, a high level of intersubject variability
is one of the major factors limiting clinical use of this
technique.
2 This variability has been recognized in
standard central field VEPs and has been attributed to many factors,
such as age and sex,
11 12 13 14 15 16 17 18 cortical convolution and
position of the calcarine fissure relative to external landmarks
(inion),
2 7 19 20 21 eccentricity of the visual field
tested,
22 conductivity of underlying tissues (bone and
subcutaneous fat thickness and blood circulation), general level of
brain activity,
23 24 25 and stimulus
conditions.
26 Although some factors (i.e., cortical
convolution) are difficult to eliminate,
2 others (such as
age, sex, scalp conductivity, and level of brain activity) when
compensated for may reduce the amplitude variation of the VEP.
Basar et al.
23 and Rahl and Basar
25 demonstrated an inverse relationship between amplitudes of the α
(8–13 Hz) and θ (4–7 Hz) components of the spontaneous EEG
activity measured immediately before stimulation, with subsequent
frontal VEP amplitudes.
23 25 They surmised that a high θ
rhythm (which is associated with drowsiness) is associated with a
suppressed VEP. When stimuli were applied only if the root mean square
value of the ongoing EEG at the lead F4 was below an individual
threshold level (so-called selective stimulation) the amplitude of the
VEP significantly increased.
With relation to sex, it has been shown that the amplitude of the VEP
is larger in women (particularly in older women) than in
men.
12 13 15 An interesting observation is that the larger
amplitude is attributed to unusually high responsiveness of the visual
system of older women to patterned stimuli
12 or to the
level of estrogen.
16
With VEP latency, a significant age effect (increasing latency with
age) has been demonstrated, but not in relation to
gender.
17 The converse holds true for VEP amplitude,
however, with no age effect being observed.
15 18 27 Dependence of amplitude variability on visual field eccentricity has
been demonstrated with the coefficient of variability (CV) of
amplitudes of the waveforms in midperipheral locations being larger
than those of the more central areas.
22
There have also been attempts to bypass the problem of between-subject
differences in cortical anatomy by inter-eye comparison (asymmetry
analysis) in the multifocal VEP interpretation.
6 7 Underlying cortical convolution and position of the visual cortex
relative to external landmarks are major contributors to intersubject
VEP variability, but they influence the signals for the two eyes of a
subject equally. This can permit the detection of unilateral changes
without reference to normal values—for example, in the case of optic
neuritis.
9 However, if there is bilateral disease, the
technique is less applicable.
In a pilot study we identified a strong correlation between background
EEG levels recorded simultaneously with multifocal VEP stimulation, and
the mean amplitude of the VEP. We surmised that the amplitudes of the
two responses were correlated because the conduction of electrical
signals across the skull, skin, subcutaneous tissue and electrodes was
altered proportionately for the two signals. Therefore, it may be
possible to use the underlying EEG levels to normalize VEP signals for
each patient, to minimize influence of the mentioned factors and thus
reduce intersubject variability. Because EEG activity is not totally
independent of visual activity (for example, α rhythm levels vary
between subjects and are suppressed by visual attention), other factors
in the raw EEG signal must be taken into consideration.
Reducing intersubject variability is crucial in the identification of
normal and pathologic results, whereas low intrasubject variability is
important for detection of progression of the disease. The purpose of
this study was to investigate variability among subjects for the
multifocal VEP, and to determine whether an EEG-based scaling algorithm
could be used to effectively reduce this variability.
One hundred normal subjects (44 men and 56 women) participated
in this study. The mean age was 58.9 ± 10.7 years (range 21–80;
men, 58.2 ± 11.1 years; women, 59.0 ± 9.9years). The study
protocol was approved by our regional ethics committee and adhered to
the tenets of the Declaration of Helsinki. Informed consent was
obtained from all subjects. All participants were examined by an
ophthalmologist. They had normal intraocular pressure, normal findings
on ophthalmoscopy, and no family history of glaucoma or retinal
dystrophy. All performed normal field tests (program 24-2; Humphrey
Instruments, San Leandro, CA), confirmed by a normal result on the
glaucoma hemifield test analysis. The inclusion criteria for the study
required corrected visual acuity of 6/9 or better and pupil diameter of
least 2.5 mm without dilation. Subjects with diabetes, previous
cataract surgery, or any other ocular disorders were excluded.
A multifocal VEP was recorded using a multifocal objective
perimeter (AccuMap; ObjectiVision Pty. Ltd., Sydney, Australia), which
simultaneously stimulates multiple sites within the visual field and
extracts corresponding VEP signals from those sites. The perimeter used
a spread-spectrum technique with families of binary sequences used to
drive the visual stimulus. Two opposite checkerboard pattern conditions
underwent pseudorandom binary exchange at each of 58 sites in the
visual field. Each input (stimulation site) was modulated in time
according to a different sequence (in contrast to m-sequences for which
the same sequence is used but shifted in time). The technique permits
computation of the resultant signal by cross-correlation of the
response evoked by the sequence stimulation with the sequence itself.
Short sequences of 4096 elements were used, which resulted in 55
seconds of recording time for each run. In further runs, different
sequences were used for the same stimulation site to reduce the
potential for cross-contamination. Results were viewed on screen after
each run and then online averaged, and the recording was terminated
when stable signals were achieved.
The visual stimulus was generated on a computer screen (22-in.
high-resolution display; Hitachi, Tokyo, Japan) with a stimulation rate
of 75 Hz. Fifty-six closely packed segments in a dart-board
configuration were used, with two additional segments located in the
nasal step region. The segments were cortically scaled with
eccentricity to stimulate approximately equal areas of cortical
(striate) surface
(Fig. 1) . The cortical scaling produced a signal of similar amplitude from each
stimulated segment. Each segment contained a checkerboard pattern (16
checks) with the size of individual checks being proportional to the
size of the segment and therefore also dependent on eccentricity. The
central area of 1° was not stimulated but was used as a fixation
monitor. Numbers of similar shape (3, 6, 8, or 9) were displayed in
random sequence, and the subject was asked to respond by pressing a
button when a particular number appeared. This ensured good
concentration throughout the recording, and the percentage of missed
and incorrect responses was calculated automatically after each run.
The luminance of the white check was 146 candelas
(cd)/m2 and the luminance of the black check was
1.1 cd/m2, producing a Michelson contrast
of 99%. The background luminance of the screen was maintained at
a mean level of 73.5 cd/m2. A dim room light was
always on.
Subjects were seated comfortably in a chair and asked to fixate on the
fixation number at the center of the stimulus pattern. The distance to
the screen was 30 cm, corresponding to a radial subtense for the
stimulus of 24°, not including the additional nasal step (32°). All
subjects had optimal refraction for near and the pupils were not
dilated. All recordings were collected using monocular stimulation.
Data were recorded using a four-channel amplifier (Grass model 15
Neurodata; Astro-Med, Inc., West Warwick, RI). The signal was amplified
100,000 times and band-pass filtered between 3 and 30 Hz. The upper
band-pass filter at 30 Hz was relatively low and outside International
Society for Clinical Electrophysiology of Vision (ISCEV) standards for
conventional VEP recording. We deliberately chose this, because it
removed high-frequency noise that can contaminate some recordings. We
have performed comparisons on the same subjects, both healthy persons
and subjects with glaucoma, and found minimal differences between the
100- and 30-Hz cutoffs other than a slight (2–3-msec) increase in
latency. Fourier analysis showed minimal high-frequency components in
the VEP above 30 Hz that have been removed by filtering (authors’
unpublished data, 1999). Although not ideal, it allowed the
system to operate in a clinical setting with the effects of muscle
noise removed and allowed use of an unshielded room.
The data-sampling rate was 450 Hz. Raw data were scanned in real time,
and noise artifacts exceeding 3 SE were excluded from the analysis.
Runs contaminated by a high level of noise were also rejected. Usually,
eight runs were recorded to provide a good signal-to-noise ratio.
VEP traces were analyzed using custom-designed software. Largest
peak-to-trough amplitudes for each wave within the interval of 60 to
180 msec were determined and compared among channels for every
stimulated segment of the visual field. The VEP waveform was most
frequently present as a single wave, simplifying identification of
peak-to-trough amplitudes. However, in some cases, a double peak could
be seen, which makes it possible that different peaks were measured in
some subjects. This may have an influence on determining latency values
(not analyzed in this article). The wave of maximal amplitude from each
point in the field was automatically selected, and a combined
topographic map was created by the software. A combined trace array was
then used for further analysis.
The intersubject CV (SD/mean) was calculated for each of the 58
segments of the visual field. The VEP amplitude was averaged over the
whole visual field (all 58 segments) for each subject, and variability
for the averaged amplitude was calculated.
To quantitatively analyze the relationship between the EEG and VEP, a
Fourier power spectrum (fast Fourier transform [FFT]) of the EEG for
each recorded channel was calculated
(Fig. 2A) . It was noted that in some subjects there was a large peak in the FFT
at approximately 8 to 10 Hz that was attributed to α rhythm
(Fig. 2B) . In some subjects there was also a strong electrocardiogram (ECG)
contribution. The ECG had previously been noted in several subjects
during real-time recording, seen as spikes that were synchronous with
the subject’s pulse. These were identified in the FFT
(Fig. 2C) . To
exclude the influence of these two components on scaling, the Fourier
power spectrum within the interval 0 to 30 Hz was fitted with a
polynomial function of the fourth order, and the integral of the fit
was calculated. Average values of the integral from all 100 subjects
were obtained for each channel.
Multifocal VEP recording provides a unique opportunity for
objective detection of visual field defects. However, high intersubject
variability due to variation in cortical anatomy caused some
investigators to conclude that it would not be useful for clinical
testing.
2 It was argued that in some locations, even
extreme damage of the visual pathway may not be reliably distinguished
from normal values.
7 This study demonstrates that using
underlying EEG amplitudes to scale an individual’s VEP response
substantially reduces this variability, including differences between
men and women. This renders the use of a normal database more reliable
when applying the multifocal VEP in the clinical detection of visual
field changes.
The intersubject variability assessed after derivation of the VEP by
cross-correlation of the raw EEG data with stimulating sequences was
evenly distributed across the visual field and very similar between the
two eyes. VEP amplitude did not correlate with age, which is in
accordance with previous reports.
15 18 27 However, there
was a well-defined difference in VEP amplitude between genders, with
the women demonstrating significantly higher amplitudes than the men.
Slight decreases in variability of gender groups compared with the
tested population as a whole indicated that some of the intersubject
VEP variability was gender based. EEG scaling of the VEP was able to
effectively remove the differences between the sexes.
The VEP amplitude was highly correlated with raw EEG data collected
during the recording. The high correlation
(
r 2 > 0.65) suggests that similar
influences account for the amplitude variability of these two
electrophysiological measures.
28 Because the VEP signal is
tiny in comparison to the EEG (approximately 1000 times) it does not
make a significant contribution to the overall EEG amplitude and is
only identified by cross-correlation techniques. We propose that
normalization of the VEP by the EEG removes the common source of
variability. We suspect that the differences between individuals are to
a large extent conductivity differences affecting transmission of the
signal from the cortex to the scalp.
The scaling applied in this study to the VEP amplitude reduced
intersubject variability greatly. An intersubject CV of the mean VEP
amplitude decreased by more than 46%, whereas individual segments of
the stimulated visual field improved variability by approximately 40%.
The scaling procedure did not affect the independence of the VEP
amplitude with age. However, the clear difference in amplitude of the
VEP between genders before scaling was eliminated by EEG-based
normalization.
It is critical to determine the components of the raw EEG signal before
applying any scaling. Some individuals have high α rhythm activity,
even when they are visually attentive. If this is included in the
scaling, the VEP response is artificially scaled down. High α rhythm
may also indicate that the patient is not concentrating and can provide
some real-time feedback to the recording technician, especially if it
appears halfway through the recording. It could also be present in
malingerers who are deliberately defocusing. Our system cannot
differentiate the cause of a high α rhythm, but it should be excluded
from any scaling algorithm used.
There are other measures that can be taken to attempt to reduce
variability. The randomly changing fixation number in the central
screen helps to keep the subject concentrating and mentally alert. It
provides a partial index of reliability and reduces the mesmerizing
effects of the multifocal display, which can cause fatigue in some
patients. Breaks between runs and the presentation of an alternative
image between runs (e.g., scenic photographs) also help in relaxation
and preventing fatigue during the test. Standardizing distance to the
screen may increase reproducibility; a tracking device has been
developed for this purpose.
The use of multichannel recording
8 reduces the great
variability between individuals that is thought to be a result of
underlying convolution of the cerebral cortex, because most dipole
orientations are covered by at least one channel. However, because of
the significant size of the area of visual field stimulated by a single
zone, the visual cortex to which this part of the visual field projects
is still not uniformly oriented (otherwise, all differences might be
rectified by differently oriented channels) but contains a
three-dimensionally curved cortex producing a signal of variable source
between subjects. Although reduction of the size of the single
stimulated zone may help to reduce the amplitude variability, it would
also lead to a reduction in the VEP amplitude and therefore to a
reduction in the signal-to-noise ratio, which would further increase
variability. Because the VEP amplitude does not seem to depend on the
age of the subjects and using EEG scaling removes variability related
to conductivity and sex, we believe that the remaining variability is
probably due to residual microanatomic differences in the cortical
convolutions of the striate cortex between subjects that cannot be
overcome by use of multichannel recording. Therefore, we still need to
derive a source-localization technique that will accurately pick up
responses from all underlying anatomic variations. Several different
approaches to this problem are currently under review.
In conclusion, by the application of EEG scaling to VEP responses, a
considerable problem in objective visual field mapping can now be
largely overcome. Interindividual variability is halved, which allows
meaningful comparisons with normal databases and increases the
sensitivity of the test to the detection of disease.
AIK is a Sydney Medical Foundation research fellow.
Submitted for publication December 19, 2000; revised March 19, 2001; accepted April 13, 2001.
Commercial relationships policy: P.
The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “
advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Corresponding author: Alexander I. Klistorner, Save Sight Institute, Sydney Eye Hospital, Macquarie Street, PO Box 1614, Sydney 2001, Australia.
[email protected]
Table 1. VEP CVs for the Whole Group and for Separate Gender Groups, before and
after EEG-Based Normalization
Table 1. VEP CVs for the Whole Group and for Separate Gender Groups, before and
after EEG-Based Normalization
| Groups | Individual Segments before Scaling | Individual Segments after Scaling | Averaged Response before Scaling | Average Response after Scaling |
| Whole Group | | | | |
| LE | 41.4 ± 5.2 | 24.6 ± 3.1 | 27.5 | 15.2 |
| RE | 42.2 ± 5.8 | 24.0 ± 3.2 | 28.5 | 14.9 |
| Female | | | | |
| LE | 39.8 ± 5.4 | 24.2 ± 2.8 | 25.8 | 15.0 |
| RE | 40.1 ± 5.2 | 24.1 ± 3.0 | 25.9 | 14.9 |
| Male | | | | |
| LE | 39.3 ± 4.8 | 24.4 ± 2.9 | 24.9 | 14.8 |
| RE | 39.8 ± 4.4 | 23.8 ± 3.2 | 25.3 | 15.0 |
The authors thank Iouri Malov and Alex Kozlovski for development of
the software program used in the analyses.
Sutter EE, Tran D. The field topography of ERG components in man. 1: The photopic luminance response. Vision Res
. 1992;32:433–466.
[CrossRef] [PubMed]Baseler HA, et al. The topography of visual evoked response properties across the visual field. Electroencephaogrl Clin Neurophysiol
. 1994;90:65–81.
[CrossRef] Klistorner AI, et al. Multifocal topographic visual evoked potential: improving objective detection of local visual field defects. Invest Ophthalmol Vis Sci
. 1998;39:937–950.
[PubMed]Graham SL, Klistorner A. Electrophysiology: a review of signal origin and applications to investigating glaucoma. Aust NZ J Ophthalmol
. 1998;26:71–85.
[CrossRef] Graham SL, et al. Objective perimetry in glaucoma-recent advances using multifocal stimuli. Surv Ophthalmol
. 1999;43(suppl 1)S199–S209.
[CrossRef] [PubMed]Graham SL, et al. Objective VEP perimetry in glaucoma: asymmetry analysis to identify early deficits. J Glaucoma
. 2000;9:10–19.
[CrossRef] [PubMed]Hood DC, et al. An interocular comparison of the multifocal VEP: a possible technique for detecting local damage to the optic nerve. Invest Ophthalmol Vis Sci
. 2000;41:1580–1587.
[PubMed]Klistorner A, Graham SL. Objective perimetry in glaucoma. Ophthalmology
. 2000;107:2283–2299.
[CrossRef] [PubMed]Hood DC, Odel JG, Zhang X. Tracking the recovery of local optic nerve function after optic neuritis: a multifocal VEP study. Invest Ophthalmol Vis Sci
. 2000;41:4032–4038.
[PubMed]Klistorner A, Graham SL, Martins A. Multifocal pattern electroretinogram does not demonstrate localised field defects in glaucoma. Doc Ophthalmol
. 2000;100:155–165.
[CrossRef] [PubMed]Allison T, et al. Developmental and aging changes in somatosensory, auditory and visual evoked potentials. Electroencephalogr Clin Neurophysiol
. 1984;58:14–24.
[CrossRef] [PubMed]La Marche JA, et al. Amplitudes of visually evoked potentials to patterned stimuli: age and sex comparisons. Electroencephalogr Clin Neurophysiol
. 1986;65:81–85.
[CrossRef] [PubMed]Celesia GG, Kaufman D, Cone S. Effects of age and sex on pattern electroretinograms and visual evoked potentials. Electroencephalogr Clin Neurophysiol
. 1987;68:161–171.
[CrossRef] [PubMed]Schoon DV, Enomoto H, Harris MP. Age-related changes in the first order Weiner kernel visual evoked potential. Doc Ophthalmol
. 1989;71:329–340.
[PubMed]Mitchell KW, Howe JW, Spencer SR. Visual evoked potentials in the older population: age and gender effects. Clin Phys Physiol Measure
. 1987;8:317–324.
[CrossRef] Kaneda Y, et al. Visual evoked potential and electroencephalogram of healthy females during the menstrual cycle. J Med Invest
. 1997;44:41–46.
[PubMed]Shaw NA. Changes in the cortical components of the visual evoked potential with age in man. Aust J Exp Biol Med Sci
. 1984;62:771–778.
[CrossRef] [PubMed]Wright CE, et al. The influence of age on the electroretinogram and visual evoked potential. Doc Ophthalmol
. 1985;59:365–384.
[CrossRef] [PubMed]Steinmetz H, Gunter F, Bernd-Ulrich M. Craniocerebral topography within the international 10–20 system. Electroencephalogr Clin Neurophysiol
. 1989;72:499–506.
[CrossRef] [PubMed]Stensaas SS, Eddington DK, Dobelle WH. The topography and variability of the primary visual cortex in man. J Neurosurg
. 1974;40:747–755.
[CrossRef] [PubMed]Hood DC, Zhang X. Multifocal ERG and VEP responses and visual field: comparing disease-related changes. Doc Ophthalmol
. 2000;100:115–137.
[CrossRef] [PubMed]Yu MZ, Brown B. Variation of topographic visually evoked potentials across the visual field. Ophthalmic Physiol Opt
. 1997;17:25–31.
[CrossRef] [PubMed]Basar E, et al. Spontaneous EEG theta activity controls frontal visual evoked potential amplitudes. Electroencephalogr Clin Neurophysiol
. 1998;108:101–109.
[CrossRef] [PubMed]Brandt ME, Jansen BH. The relationship between prestimulus-alpha amplitude and visual evoked potential amplitude. Int J Neurosci
. 1991;61:261–268.
[CrossRef] [PubMed]Rahn E, Basar E. Enhancement of visual evoked potentials by stimulation during low prestimulus EEG stages. Int J Neurosci
. 1993;72:123–136.
[CrossRef] [PubMed]Tobimatsu S, Tomoda H, Kato M. Normal variability of the amplitude and phase of steady-state VEPs. Electroencephalogr Clin Neurophysiol
. 1996;100:171–176.
[CrossRef] [PubMed]Tobimatsu S, et al. Age-related changes in pattern visual evoked potentials: differential effects of luminance, contrast and check size. Electroencephalogr Clin Neurophysiol
. 1993;88:12–19.
[CrossRef] [PubMed]Howell DC. Statistical Methods in Psychology. 1992; Duxbury Press Belmont, CA.