suggests changes in neural processing in neglect patients. Keywords: Neglect, Evoked potentials, Visual latency. Introduction. Hemineglect is a common ...
Visual Neuroscience (1994), 11, 909-918. Printed in the USA. Copyright © 1994 Cambridge University Press 0952-5238/94 $5.00 + .00
Spatial neglect is associated with increased latencies of visual evoked potentials
DONATELLA SPINELLI, 1 - 2 DAVID C. BURR, 13 AND M. CONCETTA MORRONE 3 4 'Dipartimento di Psicologia, Universita di Roma, Via dei Marsi 78, Rome, Italy Clinica S. Lucia, Rome, Italy 3 Istituto di Neurofisiologia del CNR, Pisa, Italy 4 Scuola Normale Superiore, Pisa, Italy 2
(RECEIVED October 21, 1993; ACCEPTED March 30, 1994)
Abstract
We have recorded steady-state visual evoked potentials (VEPs) from patients with vascular damage to their right brain hemispheres, some suffering from unilateral spatial neglect (n = 9), and some not (n = 7). VEPs were recorded in response to sinusoidal gratings of 0.56 cycle/deg contrast-reversed sinusoidally at temporal frequencies from 4-11 Hz. Stimuli were presented either to the left or to the right visual field, or to both. Confirming previous reports, reliable VEPs were recorded from stimuli in the left contralesional hemifield, of comparable amplitude to those of the ipsilesional hemifield and to those of both hemifields of brain damaged patients without neglect. However, analysis of apparent latency derived from phase data showed that the VEPs from the contralesional hemifield were systematically delayed by 30-40 ms compared with those of the ipsilesional hemifield, and compared with both hemifields of the nonneglect groups. This result suggests changes in neural processing in neglect patients. Keywords: Neglect, Evoked potentials, Visual latency
For example, the cases showing neglect in mental images (mentioned above) show that neglect phenomena can be present independently of sensory information. Other evidence shows that symmetry information used for figure-ground separation can be acquired "pre-attentively" to influence perception, without subjects being directly aware of the images in the neglected hemifield (Driver et al., 1992). Finally, two recent studies have shown that visual evoked potential (VEPs) can be recorded from the contralesioned hemifields in neglect patients (Vallar et al., 1991; Viggiano et al., 1994), suggesting intact processing in primary visual areas. In this study, we investigate further the visual evoked response of neglect patients, using somewhat more subtle techniques than hitherto applied. The results show that although the visual responses to contralesional stimulation are similar in amplitude to those of ipsilateral stimulation, there are clear phase differences, implying differences in response latency. These results would suggest that the neglect syndrome is accompanied by measurable changes in primary processing of sensory information. A preliminary report of these findings has been published in abstract form (Spinelli et al., 1993).
Introduction
Hemineglect is a common neurological syndrome in patients with cerebral lesions the right hemisphere. The clinical symptoms are a tendency to ignore all sensory information falling in contralesional (left) space. Severe cases will fail to respond to a speaker in the left side of the room; will not touch food on the left side of their plate, complete only the left half of a crossword; and even split words in two, reading only the right half. Some patients also exhibit hemineglect to mental images, reporting, for example, the contents of only one half of a wellknown public square when imagined from one vantage point, and the other half when imagined from the opposite vantage point (Bisiach & Luzzatti, 1978; Guariglia et al., 1993). There are several theories about spatial neglect (summarized and discussed in the excellent reviews of Bisiach & Vallar, 1988, and Rizzolatti & Gallese, 1988). The various theories fall loosely into two categories, involving "low-level" and "high level" explanations. The low-level approach seeks to implicate reduced or disordered processing of sensory information, or impairment of sensorimotor integration, while high-level explanations invoke concepts such as attention and spatial representation at higher cortical levels. There is little evidence to support an extreme formulation of a theory attributing neglect to impaired sensory processing.
Methods Stimuli Steady-state visual evoked potentials (VEPs) were recorded in response to stimulation by vertical gratings modulated sinusoi-
Reprint requests to: Professor David Burr, Istituto di Neurofisiologia del CNR, Via S. Zeno 51, Pisa, Italy.
909
D. Spinelli, D.C. Burr, andM.C. Morrone
910
dally in space and time. The gratings were 150 cd/m 2 mean luminance, 32% contrast, and 0.56 cycle/deg spatial frequency, reversed sinusoidally in contrast at variable temporal frequencies (fundamental 4-11 Hz). They were generated on a Joyce Electronics monitor at 100 frames/s by means of a Tessa-II waveform generator (Burr & Eikelboom, 1987). The oscilloscope screen was masked by white card to reveal two rectangular patches 9 cm wide by 23 cm high, separated by a 2 x 23 cm central strip with a clear fixation point in the middle. When viewed from 80 cm, each rectangle subtended 6.4 by 16.4 deg, distanced by 0.7 deg from the fixation point. The two exposed rectangles could be separately occluded, to study three experimental conditions: stimulation of only the left hemifield (referred to as LVF), only the right hemifield (RVF), or combined stimulation of both fields (LVF + RVF). Observers were instructed to maintain steady fixation of the central fixation point. An experimenter monitored eye movements visually, and interrupted data collection (with a hand held switch) if fixation deviated. Electrophysiological techniques VEPs were recorded differentially with Ag-AgCl electrodes placed 2 cm above the inion (Oz) and at the vertex (Cz) with ground half way between (P,). VEP signals were amplified (50,000-fold), band-pass filtered between 1-100 Hz (6 dB/oct), and averaged by computer after artifact rejection (12-bit resolution, 64 samples per period). The computer rejected single sweeps over a threshold voltage, to minimize contamination by eye blinks or gross eye movements. A PC computer averaged the EEG in synchrony with the stimulus contrast reversal rate, to yield steady-state VEPs (Campbell & Maffei, 1970; see Porciatti et al., 1992 and Morrone et al., 1993). The computer performed an on-line discrete Fourier analysis to estimate the amplitude and phase of the secondharmonic modulation (the principle modulation frequency). The program also averaged the signals asynchronously at 1.1 times the temporal frequency of the stimulus to give an estimate of background noise, that was used to estimate signal-to-noise ratios (illustrated in Fig. 1). A separate estimate of signal reliability was obtained from the two-dimensional scatter of partial averages of the signal (see Fig. 1). For further details about techniques of estimating reliability of steady-state VEPs, see Victor and Mast (1991). Response latencies were estimated from the steady-state VEPs by measuring phase as a function of temporal frequency, and calculating the slope of the curve (see Regan, 1966; Spekreijse et al., 1977; Porciatti et al., 1992). The slope was calculated by least-squares fit, after weighting each data point by its signal-to-noise ratio and by the inverse of the standard error of the mean of the phase variability (see Fig. 1). Estimating latency from phase poses a problem, as there exist an infinite set of values, separated by 2ir, that are mathematically identical. To circumvent this problem, one must assume that the phases advance or retard in an orderly fashion with temporal frequency, and add or subtract multiples of 2ir to the data to produce maximum orderliness (as determined by leastsquared regression). In any event, to be certain that the estimates of latency differences were correct, we estimated them with several different techniques (described in more detail in the Results section).
50 ms Fig. 1. Examples of averaged steady-state VEPs, recorded from a normal control (A, B) and from a patient with spatial neglect (C, D), in response to stimulation with a 0.5 cycle/deg grating modulated at 7 Hz. Records A and C show responses to stimulation of the right visual field and B and D to stimulation of the left (contralesioned) visual field. The continuous lines show the averages at the frequency of stimulation, and the dashed lines the averages at 1.1 times this frequency (to give an indication of background noise levels). All traces show a clear modulation at twice the stimulation frequency, well above the levels of noise modulation (dashed lines). The plots on the right show the amplitude (distance from origin) and phase (argument) of subsets of the averaged response (individual packets of 20 sums). The degree of scatter gives an indication of the consistency of the responses.
Subjects VEPs were recorded from 23 patients with right hemisphere vascular damage (average age 64 ± 13) and from 16 healthy young adults (average age 25 ± 3). Seven of the patients were eliminated from analysis (on the grounds described below). Of the remaining 16 patients, nine displayed evidence of spatial neglect, according to standard neuropsychological tests [two cancellation tests, one reading test, and the Wundt-Jastrow illusion test (Pizzamiglio et al., 1989)]. A brief description of all patients, including age, position location of lesion, and degree of neglect is given in Table 1. The patients without the neglect syndrome provided the principle control for the neglect patients, as they were similar in age. Informed consent was obtained from all subjects after the nature of the technique was fully explained.
911
VEPs latency from patients with spatial neglect Table 1. Summary of neurological information of the 16 patients used in this study (all suffering from vascular damage)" Case
Sex
Age
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
F M F F F F M F M M M M F M M M
68 70 68 79 63 63 78 70 45 74 61 69 31 79 64 44
Lesions
Time from onset (months)
R R RTP R RTP RFP RFO R T + bg Rbg Rbg RTP R RFT + bg R capsula ROP Rbg + ext. capsula
30 7 4 4 10 11 3 2 4 6 3 3 3 1.5 3 19
Neglect
RVF latency (ms)
LVF latency (ms)
135 124 150 92 115 148 88 89 94 168 112 82 125 135 69 88
169 146 161 149 145 169 124 135 114 129 99 68 93 118 74 88
a Sex —F: female, and M: male. Lesions —R: right, T: temporal, P: parietal, O: occipital, F: frontal, and bg: basal ganglion. For cases indicated only by R there was no CT scan available, so the lesion was assessed by standard neurological tests. The last two columns show the estimates of VEP latency obtained in this study from Fig. 5.
Selection of subjects An obvious prerequisite for studying VEPs (in patients) is that the patient has a reliable visual evoked response to the stimuli used: to include in the averaged data patients with no response under any conditions data would only increase the variability of the data, diluting real differences without adding any useful information. On the other hand, it is important to have objective criteria for rejection, so that the rejection process itself does not bias the results. The criteria for rejection of patients was to consider the reliability of the evoked potentials recorded on stimulation with the ipsilesioned (normal) hemifield. Here we evaluated the signal-to-noise ratio of the potentials in each condition (the ratio of VEP amplitude evaluated with synchronous averaging to that evaluated by asynchronous averaging: see Fig. 1). If the signalto-noise ratio, averaged over all temporal frequencies, was less than 1.5 (implying less than 50% signal above the noise level), the results of that patient were excluded from further analysis. In practice, three patients were rejected from the neglect group, and four patients without neglect were rejected. Table 1 describes only patients that survived the signal-to-noise criterion, and hence were included in the analysis.
Results
Steady-state VEPs As detailed in the Methods section, steady-state VEPs were measured for the three experimental groups: patients with right brain damage that exhibit the neglect syndrome (indicated "NEGLECT" on all graphs), patients with right brain damage without neglect (RBD), and healthy young subjects (NORMAL). Fig. 1 shows examples of VEP records for a normal subject (upper traces) and for a neglect patient (lower traces). The traces
show averages of 200 cycles of contrast reversal, with stimulation of either the right or left visual fields (illustrated with sketch). All traces show two distinct humps, indicating modulation at each contrast reversal, that occur at twice the frequency of stimulation (second-harmonic modulation). For the neglect patient, the records are of comparable amplitude (slightly higher amplitude for the neglected hemifield in this particular case), showing that the strength of the response was similar with stimulation of left or right hemifields. The dashed lines show the level of asynchronous noise, averaged at 1.1 times the frequency of stimulation. This gives a level of the baseline noise at a frequency near that of interest, which helps to evaluate the reliability of the signal. A further indication of reliability is given by the polar plots on the right of the averaged records. These show the amplitude and phase of the second harmonic (see Methods) of individual twenty-sum "packets" of the VEP. The degree to which they all fall together, with similar phase and amplitude, is a further indication of the reliability of the VEP, and was used to estimate the standard error of the mean (see also Victor & Mast, 1991). Note that in these selected examples, the average phase of the second harmonic (the center of the scatter of points) of the neglect patient (but not the normal subject) is different for right and left hemifield stimulation, although the stimulus conditions (contrast, spatial frequency, temporal frequency, etc.) were identical in both cases. This was in fact observed in many cases, and will be pursued in some detail in later sections.
Amplitude as a function of temporal frequency For the three experimental groups, VEPs were measured in response to stimulation of the left visual field (filled circles in all graphs), the right visual field (open circles), and both fields (open squares). All VEPs were analyzed to yield the amplitude
D. Spinelli, D.C. Burr, andM.C. Morrone
912
damaged patients to show higher (or lower) levels of baseline noise at the frequencies measured here. Fig. 3 reports the same data, but averaged after response normalization, to minimize biases that may result from variation in average VEP amplitude (possibly due to strictly physical factors). The response curves were normalized by dividing the amplitudes of each subject by the sum of amplitudes recorded with right hemifield stimulation, and multiplying by 100 (normalizing the total amplitudes from right hemifield to 100). The normalized responses were then averaged over subjects. This process abolishes any global scaling effects in amplitude while preserving, and perhaps highlighting, relative differences. Fig. 3 shows the averaged normalized amplitudes for the three groups, displayed with the same format and symbol type as in Fig. 2. The error bars show ± 1 standard error, calculated from the variation between subjects. The dotted curves show the average asynchronous noise levels. The basic trend of the data is the same as for Fig. 2, showing that the pattern of results does not depend critically on the averaging strategy. The normalized curves show band-pass functions for normal, and lowpass functions for both patient groups, with the slight tendency for VEPs of the contralesional hemifield of the neglect group to be more low-pass and have a lower temporal acuity than those of both control groups.
and phase of the second harmonic of the response. In this section, we discuss the amplitude measurements. Fig. 2 illustrates how average VEP amplitude varied with temporal frequency for the three groups and for the three experimental conditions. In this representation, the VEP amplitudes have been averaged over all subjects within the group, without normalization or weighting. There is a strong and reliable response over this range for both the neglect and nonneglect patients, although the amplitudes of the normal subjects are somewhat higher than those of both patient groups. The response of the normal subjects tends to be band-pass (showing attenuation for both higher and lower temporal frequencies) while the response of both patient groups is low-pass, with no attenuation at low temporal frequencies. At most temporal frequencies, the response of the neglect group was similar to the nonneglect group, but there are some subtle differences. For example, at high temporal frequencies, the response of the neglect group was lower for the left than for the right visual field, suggesting stronger attenuation at higher temporal frequencies for the neglected hemifield. For the normal and nonneglect groups, VEPs recorded to combined presentation of left and right fields were about twice the amplitude of those recorded to each field separately, suggesting simple magnitude summation of the activity evoked by each hemifield. However, with the neglect group, the amplitudes to combined stimulation were similar to those recorded separately. This will be taken up in more detail in a later section after response phase has been discussed. The lines without symbols indicate the average levels of asynchronous noise for the three conditions. Note that the noise levels are about the same for the three conditions, and for the three experimental groups. There is no obvious tendency for the brain
Response phase with temporal frequency A complete description of a steady-state evoked potentials includes both amplitude and phase information. Whereas the amplitude is an index of the strength of the response, the phase is related to the response latency. Although latency cannot be recovered completely from a single measure of phase, it is pos-
NORMALS
RBD
3 -i
W
3
o J
NEGLECT
-[
3-1
2 -
2-
1 -
1 -
H
$
1 H
j
6
8
T
|
10
•
1
12
2
=
^
I
'
4
^
_
I
'
6
-
^ I
8
10
12
2
4
6
8
10
TEMPORAL FREQUENCY (Hz) Fig. 2. Average VEP amplitudes for the three groups of subjects, as a function of fundamental temporal frequency (reversal rate is twice this frequency). The open squares refer to stimulation of both hemifields (stimuli 6.4 deg wide by 16 deg high, positioned either side of a 1.4 by 16 deg blank patch), the closed circles to stimulation of only the left (contralesional) visual field, and the open symbols to stimulation of the right (ipsilesional) visual field. The continuous, dashed, and dotted lines show the average asynchronous noise for dual, left, and right hemifield stimulation. The average standard error of the amplitudes, estimated from the scatter of the VEP (see polar plots of Fig. 1), were 0.2-0.3 jiV, very similar to the average noise estimates. The average standard deviation of the inter subject scatter was 0.9 ^V for both groups of patients, and 1.1 yiW for the normals.
12
913
VEPs latency from patients with spatial neglect
NORMALS
RBD
NEGLECT
40 -
40 -i
40 -i
20-
20 -
20 -
a
10
12
8
10
12
10
12
TEMPORAL FREQUENCY (Hz) Fig. 3. VEP amplitudes, averaged after normalization: the amplitudes of each subject were normalized so that the sum of amplitudes with right-hemifield stimulation totaled 100. The averages are reported together with ±1 s.E. As with Fig. 2, the open squares refer to stimulation of both hemifields, the closed circles to stimulation of only the left hemifield, and the open symbols to stimulation of the right hemifield. The dashed line shows the average noise for the three conditions.
sible to obtain an estimate of "apparent latency" from the rate at which phase decreases with temporal frequency (see Porciatti et al., 1992 and the Methods section for details). Fig. 4 plots the averaged phase plots for the three conditions and for the three subject groups. All of the phase curves show
NORMALS
a steady, linear decrease with temporal frequency (as is generally observed), and can all be well fit by linear regression (r > 0.9). The three curves are virtually superimposed for the normal group, but they are quite separated for the neglect group: the curve for the left visual field is much steeper than that for
RBD
NEGLECT
6 -I
6 -i
6 -i
4
4 -
4 -
2-
2 -
2 -
on Z
~
a OS
on
6
8
10
12
2
4
6
8
10
12
2
4
6
8
10
TEMPORAL FREQUENCY (Hz) Fig. 4. Phase as a function of temporal frequency (averaged over subjects), for the various conditions. The symbols are the same as those of Figs. 2 and 3 (open squares: dual fields; open circles: right field; and filled circles: left field). The slope of the best-fitting linear regression gives an estimate of response latency. The estimates for the latencies in response to right (contralesional), left (ipsilesional), and dual stimulation were, respectively, 114, 113, and 116 ms for the normal group; 94, 104, and 110 ms for the nonneglect group; and 148, 132, and 106 ms for the neglect group.
12
914
D. Spinelli, D. C. Burr, and M. C. Morrone
the right visual field, with the dual-field curve falling in between. For the nonneglect group the situation is inverted, with the leftfield curve being slightly shallower than the other two. The slope of the curves gives a direct estimate of the apparent latency of the VEP, an estimate that has been shown to correspond closely to that of the P100 in transient VEPs, at least for normal young subjects (Spekreijse et al., 1977; Riemslag et al., 1982). With the normal controls, all three latency estimates were very similar (114, 113, and 116 ms for left, right, and dual-field stimulation, respectively). The estimates of curve slope for the neglect patients are 148 ms for left-field stimulation against 106 ms for right-field stimulation, 42 ms longer. Stimulation with both fields produced an intermediate latency (132 ms). With the nonneglect patients the situation was inverted, with latencies from left stimulation being slightly less than for right stimulation (94 cf. 110 ms, with 104 ms for dual stimulation). Notice that for the neglect patients, there is a difference not only in the slope of the curves but also in the intercept at 0 Hz. This would imply not only a difference in latency, but also a phase shift. However, we must treat this conclusion with particular caution, as the technique requires addition of multiples of lit to create the linear dependence of phase on temporal frequency (see Methods). While one can be reasonably confident that the relative multiples of 2ir are correct in predicting the slope (also confirmed by other techniques, described below), the absolute values are more difficult to judge. If the whole curve of a single patient were displaced by 2w (without affecting its slope), it would create a difference in the intercept of the averaged curve. It would seem premature to speculate on the difference of intercepts at this stage. The technique used to obtain latency estimates from averaged phase data can also be applied to individual subjects. Estimates of response latency for left- and right-hemifield stimulation for individual subjects are shown in the correlograms
of Fig. 5. The dashed line indicates equal latency for left and right fields. The latencies for the normal subjects are clustered in a fairly tight group, distributed roughly equally on either side of the equilatency line. However, with the patients, all those with the neglect syndrome (filled circles) fell to the right of the equilatency line (showing longer latencies for left-field stimulation), while the latencies of the nonneglect patients (open circles) fell much closer to the diagonal, with some falling in the opposite quadrant. The latencies for each patient are also reported in Table 1. The average of the individual estimates are indicated by the arrows near the axes, with error bars showing ±1 S.E. of variation between subjects. Despite differences in the averaging strategy, the estimates of average latency are very similar to those of Fig. 4, with the major effect being a difference in latency between left and right visual field stimulation, about 31 ms with this procedure. This greatly increases our confidence in the data. As mentioned in the Methods section, evaluation of phase slope may be a delicate procedure, given the necessity to add multiples of 2TT the phase measures. For this reason, we performed an independent estimate of relative latency for each subject, based on the difference in the phase of the response for left- and right-field stimulation. For each subject, the phase with right-field stimulation was subtracted from that with left-field stimulation, and plotted as a function of temporal frequency. Again, the slope of this curve is an estimate of the difference in response latency to the two hemifields. Fig. 6 plots the direct estimates of relative latency (taken from the difference in latencies of Fig. 5) against the estimates obtained from the differences in phase. The correlation between the estimates of the two techniques is good (r = 0.91), giving us confidence in the technique. It is obvious from inspection of Fig. 5 that the relationship between latency difference and the neglect syndrome is statistically significant: the distributions are virtually nonoverlapping.
NORMALS 200
PATIENTS 200 -i
i
150 -
150 --
3° /"
CO
'
100 - I
100 -
o ,/
w E-
3
H H
0,-1
50 -
50 KX
1 50
100
150
200
50
100
150
200
LVF LATENCIES (ms) Fig. 5. Scatter plot of latencies from stimulation of right and left hemifields of individual subjects. The bars indicate ± 1 S.E., estimated by repeatedly calculating latency (1000 times) with only 70% of data points (selected at random), and estimating standard error from the spread in the partial calculations. The arrows near the axes indicate group averages, with error bars showing ±1 S.E. of the variation between subjects. The trend for the neglect group to fall to the right of the equilatency line is observed in all patients.
915
VEPs latency from patients with spatial neglect 80 -i
