Abstract
purpose. A recent study has demonstrated that some people with migraine display impairments of intermediate stages of motion and form processing. Deficits were identified by using tasks that required that local stimulus attributes be integrated into global percepts. Neurons capable of global processing of form and motion are known to be present in extrastriate cortical areas V4 and V5, respectively. It is not clear from the literature whether deficits of global processing in migraineurs are likely to arise from reduced input to extrastriate cortex from primary visual cortex (V1). The purpose of the study was to compare presumed measures of V1 performance (vernier acuity) to measures of global form and motion perception in migraineurs.
methods. Thirty migraineurs (17 with aura, and 13 without) and 20 age-matched nonheadache control subjects participated. Intermediate level motion and form perceptions were measured using global dot motion stimuli and Glass patterns, respectively. Vernier stimuli were broad vertical bars composed of small dot elements. Both a static luminance stimulus and a motion defined form vernier stimulus were used.
results. Mean migraine and control group performance were not significantly different for either vernier task (static: t (48) = 0.39, P = 0.70; motion: t (48) = 0.29, P = 0.77). Mean migraine group performance was significantly worse than in control subjects for both the global form (t (48) = 2.06, P = 0.04) and global motion (t (48) = 2.87, P < 0.01) tasks.
conclusions. On average, migraineurs demonstrate abnormalities of intermediate stages of both motion and form processing. These abnormalities do not appear to arise from dropout of performance at V1, as vernier acuity was normal in the same individuals.
Migraine is a common neurologic condition that is often associated with visual symptoms. These visual symptoms may include visual aura or more subtle visual symptoms such as blur. As symptomatology indicates that the visual pathways are involved in migrainous pathophysiology, substantial research has been conducted exploring visual processing in migraineurs (for example, Refs.
1 2 3 4 5 6 7 8 9 ). In a recent study, we identified deficits in intermediate level processing of both form and motion stimuli in individuals with migraine during the period between their attacks.
2 The ability to integrate local orientation information into a global form percept and the ability to integrate local motion information into a global motion percept were measured. Current understanding of the hierarchical organization of the human visual system indicates that V1 predominantly processes local stimulus attributes, whereas neurons in extrastriate visual areas V5 and V4 are able to integrate local information into global motion and form percepts, respectively.
10 11 12 13 There is also recent evidence for additional processing of global motion in area V3A.
14 Hence, the observation that some migraineurs have an impaired ability to perform visual tasks that require global integration may imply aberrant neural processing in the extrastriate visual cortices.
2 Recent studies using transcranial magnetic stimulation (TMS) also suggest differences in neural processing in extrastriate cortex of people who experience migraine.
15 16
An alternate explanation for the impaired ability of migraine groups on the global motion and global form tasks is that it arises as an upstream effect of dysfunction earlier in the visual pathways. An obvious place to look for such dysfunction is V1. Migrainous visual aura is retinotopic and homonymous consistent with a cortical locus
17 and transient changes in regional cerebral blood flow in area V1 have been measured during the aura phase of migraine patients.
18 19 20 21 Numerous studies have used visual psychophysics to explore V1 function at times between migraine events with variable results. Differences in reaction time to several low-level visual tasks have been noted
9 ; however, these findings were not replicated by an alternate research group.
22 Migraineurs have demonstrated enhanced sensitivity to high-contrast masking patterns,
3 6 and differences in pattern adaptation,
7 yet no difference has been found for the ability to discriminate orientation
8 or motion direction.
5 Given the differing results of these previous studies, we cannot readily assume that global motion and global form deficits in migraineurs are the result of reduced input from V1 to higher-order processing.
The purpose of the present study was to use visual psychophysics to compare presumed measures of cortical function in the motion and form processing pathways at the level of V1 and extrastriate cortex in the same individuals with migraine. Performance was compared to a group of nonheadache control subjects. In an attempt to compare form and motion processing fairly, we chose visual tasks where stimuli could be created that were as similar as possible for assessment of the two visual processing streams. For assessment of V1, we measured vernier thresholds. Vernier performance is understood to involve the cortical integration of spatial information,
23 24 and static luminance and motion defined stimuli can be created to differentially assess the spatial localization ability of the form and motion pathways.
25 26 To assess intermediate levels of motion processing we used limited lifetime global dot motion stimuli similar to those we have used previously.
2 4 Global form perception was assessed by measuring thresholds for the perception of global structure in Glass patterns.
27 These global form and motion stimuli are very similar, enabling intermediate form and motion processing to be assessed in an analogous fashion.
Thirty migraineurs and 20 nonheadache control subjects participated. Migraineurs had to have experienced migraines that met the classification criteria of the International Headache Society for either migraine with aura (MA) or migraine without aura (MO).
28 Nonheadache control subjects had to be free of regular headaches and to have never experienced a migraine. Psychophysical observation experience was approximately equivalent (minimal) between the two groups.
Seventeen of the migraineurs were classified as experiencing MA (age range, 19–42 years; median 28), whereas the remaining 13 migraineurs were classified as experiencing MO (age range, 18–43 years, median 22). The age of the control subjects ranged from 18 to 47 years (median, 26). There was no significant difference between the median ages of the three groups (Kruskal-Wallis one-way ANOVA on ranks: H(2) = 0.472, P = 0.79). All MA subjects experienced visual aura as part of their aura symptomatology.
Migraine severity was assessed using the Migraine Disability Assessment (MIDAS) questionnaire,
29 a self-report measure that assesses migraine disability by the number of days that subjects are prevented from completing day-to-day activities due to headaches. Migraineurs also retrospectively reported their age at first migraine and the number of attacks experienced during the past 12 months. The migraine characteristics of the subjects are shown in
Table 1 . Migraineurs were excluded from the study if they were taking preventative migraine pharmacotherapy.
All subjects met the following visual and ocular health criteria: best corrected visual acuity of 6/7.5 or better, spherical refractive errors between +5.00 and −5.00 D and astigmatic refractive errors between +2.00 and −2.00 D, normal findings in anterior eye and ophthalmoscopic examinations, no evidence of glaucoma, and no history of diabetes or other systemic disease known to affect ocular function with the exception of migraine. Participants were not permitted to be taking any medications known to affect visual field sensitivity, contrast sensitivity, or other ocular function.
All subjects provided written informed consent in accordance with a protocol approved by the University of Western Australia Human Research Ethics Committee and in agreement with the tenets of the Declaration of Helsinki.
All stimuli were presented on a gamma-corrected, 21-in., gray-scale monitor (Model GD-402; Phillips, Eindhoven, The Netherlands). The refresh rate was 100 Hz for the global form and motion tasks. The refresh rate was reduced to 70 Hz to enable higher spatial resolution for the vernier acuity tasks. Experiments were conducted with custom software written in a commercial program (MatLab 6.0; The MathWorks, Natick, MA) and displayed using a graphics card (VSG 2/5 graphics card; Cambridge Research Systems, Kent, UK). Subjects were seated at a chin rest and viewed the monitor monocularly from 75 cm for the global form and motion tasks, and from 10 m (via a front surface mirror) for the vernier acuity tasks. If both eyes met the refractive and visual acuity inclusion criteria for the study, subjects were asked to choose which eye they preferred to have tested. Refractive correction appropriate for the viewing distances was provided where required. The appropriate refractive correction was determined by subjective refraction and measurements of accommodative status by a registered optometrist (AMM). Subjects wore their own spectacles if suitable or alternatively wore refractive correction in a trial frame. Migraineurs were tested at least 4 days after migraine, to minimize the possible influence of transient postmigraine fatigue or nausea. Subjects were asked to report if they experienced a migraine in the first 72 hours after testing. No reports of posttesting migraine were received. All tests were completed within a single experimental session of approximately 1.5 hours’ duration.
Vernier acuity was measured using two-bar vertical vernier stimuli, which were defined either by static luminance contrast dots
(Fig. 1A)or by the relative motion of randomly placed dots
(Fig. 1B) . The static luminance-defined stimuli were composed of small dot elements (17 sec arc-square pixels), which were randomly placed within the bars. The bars were 10 min arc wide and 15 min arc high and were separated vertically by 4 min arc. The dots were white (100 cd/m
2) and were presented on a black (0.5 cd/m
2) background. For the motion-defined stimuli, both the bars and the background were covered with random dots of equivalent density (4593 dots per deg
2) throughout a central 1.23° square patch. A relatively high dot density was chosen, as motion-defined vernier acuity performance has been shown not to be limited by sampling for dot densities of this magnitude.
25 When the dots were stationary, the vernier target was invisible. The target was made visible by the motion of the dots within the bars relative to that of the background. The background dots moved upward at 1.25 deg/sec, while the dots within the bars moved downward at the same speed. The motion was created by the use of an eight-frame motion sequence, in which each frame was displayed for 57 ms, resulting in the total stimulus duration being 456 ms for the motion-defined vernier targets. The stimulus duration was identical for the static vernier targets. The density of the dots was equivalent for the static and motion-defined vernier stimuli; hence, the static stimulus is simply the vertical bars from the motion-defined case displayed without motion. For each stimulus presentation, the random dot texture was recreated to avoid any systematic influence on performance caused by the positioning of the random dot elements. We recognize that the static stimulus was broader and of lower perceived contrast than would provide optimal vernier acuity thresholds.
25 We wanted to maintain as much similarity as possible between the static luminance and motion-defined vernier stimuli, to ensure that if the migraine group displayed deficits on one of the vernier tasks alone, that performance could not have been influenced by simple differences in stimulus attributes such as dot density.
Thresholds were determined with a two-interval, forced-choice (2IFC) procedure. In the test interval, a vernier target was shown with a variable degree of horizontal displacement between the upper and lower bars, whereas the control interval showed a target with no horizontal offset. Subjects were required to choose the interval in which the stimuli were horizontally displaced. The offset was randomly chosen on each presentation to be either left or right. The amount of horizontal displacement was varied according to a staircase strategy, where the offset was reduced if three consecutive correct responses were made and increased for every incorrect response. This three-down, one-up design results in the staircase converging on the 79% correct performance level.
30 Two staircases were interleaved, and each started with an offset of 2 min arc for the static condition and 6 min arc for the motion condition. The initial staircase step size was 4 pixels (each pixel was 5.7 sec arc when viewed at 10 meters), which was halved on the first two reversals. The staircases terminated after six reversals, with the result of each staircase being determined as the average of the final four reversals. The mean of the two interleaved thresholds was taken as the threshold estimate for each task.
Global form perception was measured with Glass pattern stimuli as shown in
Figure 2 . These patterns are created by dot pairs (separated by 9 min arc) placed randomly within a 10° circular aperture. The dots were white (100 cd/m2), square pixels of 4.5 min arc diameter, and were presented on a black background (0.5 cd/m2). The orientation of the pairs was perpendicular to the center of the image, which creates a “concentric” pattern.
Figure 2Ashows a pattern in which 100% of the dot pairs are aligned in this concentric fashion. To measure global form coherence thresholds, the patterns were degraded by replacing a percentage of the signal pairs with two randomly positioned noise dots.
Figure 2Bshows an example of a 50% coherence Glass stimulus. Each pattern had a total of 200 dots (100 potential dot pairs). In
Figure 2A , all 200 dots are paired, whereas in
Figure 2B , there are 50 dot pairs and 100 randomly placed dots. Noise Glass stimuli were also created for use in a 2IFC procedure. The noise stimuli had the same number of dot pairs as the signal stimuli (also separated by 9 min arc), but instead of the pairs being arranged concentrically, the orientation of the noise pairs was random. The remainder of dots in the noise stimuli were placed randomly. This type of Glass pattern construction has been used previously to explore global form perception in both normal and clinical groups.
31 32 33
Figure 3provides a schematic representation of the global motion stimulus for a 10% motion coherence threshold (10% of the dots moving in the signal direction, which was downward). The stimulus was composed of 100 dots placed randomly within a 10° circular aperture. The dots were the same size and luminance as those used to create the global form stimuli. To create the motion percept, an eight-frame motion sequence was shown, in which each frame was displayed for 50 ms with no interframe interval. For each stimulus presentation, a percentage of the dots were chosen to move in the signal direction, while the remainder of the dots moved in random directions. All dots were displaced spatially by 9 min arc per frame (creating an effective velocity of 3 deg/sec). Signal dots were randomly chosen for each frame transition. This aspect of stimulus design minimizes the utility of tracking local motion cues by following individual dots from frame to frame, provided that signal levels are low. Consequently, global integration of local motion information is necessary to determine the coherent motion direction at normal threshold levels.
Thresholds were determined in a 2IFC procedure for both global motion and global form tasks. In the test interval, a stimulus containing a coherent signal was shown, whereas the control interval displayed a noise stimulus. Subjects were required to identify which interval contained the signal stimulus (concentric structure for the global form task, downward motion for the global motion task). The proportion of signal was varied according to a three-down, one-up staircase procedure, as for the vernier tasks. Dual interleaved staircases were used with initial signal strength of 50 dots (pairs for the global form task). The staircase step size was initially 4 dots and was halved on the first two reversals. All other aspects of the staircase procedure, including the averaging, were the same as for the vernier task.
The group mean (±SD) vernier alignment thresholds are shown in
Figure 4 , along with all individual data. The distributions of both control and migraine group data did not significantly depart from a normal distribution (Kolmogrov-Smirnov test,
P > 0.05). There was no significant difference in performance between the MA and MO groups on either task, and so all migraine data were pooled for comparison with control subjects (static luminance defined task:
t (28) = 0.63,
P = 0.53; motion defined task:
t (28) = −0.53,
P = 0.63). There was no significant difference in performance between migraine and control groups for either the static vernier task (
t (48) = 0.39,
P = 0.70) or motion-defined vernier task (
t (4 8) = 0.29,
P = 0.77). Motion defined vernier thresholds were substantially higher than those for the static bars and were consistent with threshold estimates reported previously in the literature.
25 The static vernier acuity thresholds were higher than typical reports for line stimuli. Alignment thresholds have been shown to increase with decreasing stimulus spatial frequency,
34 hence higher thresholds are expected for our wide degraded luminance bars.
Figure 5shows migraine and control group mean performance on the global form and global motion coherence tasks. All individual data are shown and the MA and MO groups are identified separately. For both tasks, the distributions of the data for all groups did not depart significantly from that of a normal distribution (Kolmogrov-Smirnov test,
P > 0.05). There was no significant difference in thresholds between the MO and MA groups on either task and so all the migraine data were pooled for comparison with nonheadache control subjects (form:
t (28) = −0.72,
P = 0.48); motion (
t (28) = −0.63,
P = 0.53). The mean migraine group threshold was significantly elevated when compared with that of control subjects for both the global form (
t (48) = 2.08,
P = 0.04) and global motion (
t (48) = 2.87,
P < 0.01) tasks. Global motion coherence thresholds were not significantly correlated with global form coherence thresholds (Pearson product moment correlation coefficient:
r 2 = −0.13,
P = 0.49).
Correlation coefficients were calculated to explore for possible relationships between the self-reported migraine characteristics and performance on the global form and global motion tasks. There was no significant correlation between the global motion or form coherence thresholds and any of the following migraine characteristics: years of migraine, number of migraines in the past 12 months, duration since last episode, although a weak negative correlation approached significance between global form thresholds and years of migraine (
P = 0.06). Correlation coefficients and corresponding probabilities are shown in
Table 2 .
In this study, migraine group mean thresholds were higher for global motion and global form perception than those of nonheadache control subjects. This finding is consistent with previous studies that have identified global motion
1 2 4 5 35 and global form
2 perception deficits in migraine. In the same subjects, vernier thresholds did not differ from those of control subjects, for either a static luminance vernier target or a motion-defined vernier target. The normal spatial localization performance suggests that the performance reduction for the global tasks did not arise from a direct dropout of neural function at V1.
23
There are several possible explanations for the reduced performance on the global tasks. These include the possibility that V1 function is normal in migraineurs and that deficits arise in extrastriate processing areas. It is difficult to reconcile this theory with previous studies that report aberrant neural responsiveness at V1 measured with electrophysiology or TMS (for review, see Refs.
36 37 38 ). An alternate explanation is that anomalous processing is indeed present at V1, but that it does not manifest as a deficit in performance on the vernier tasks. Reports of previous psychophysical studies of presumed V1 function in migraineurs have been variable, with differences in performance relative to nonheadache control subjects reported in some studies
3 6 7 but not in others.
5 8 22 Perhaps the findings of the present study can be reconciled with those of previous studies if there were aspects of the stimulus design that increased the likelihood that deficits would be found with the global motion and global form tasks but not the vernier tasks.
The global motion and global form stimuli were composed of random dots, and a proportion of the dots were signal and the remainder noise. In normal observers, the ability to integrate global motion yields a signal-to-noise ratio that is constant at threshold in low-density patterns.
39 Similarly, detection of coherence in Glass patterns by normal observers also depends on the signal-to-noise ratio; however, the threshold increases more slowly with elevations in noise than in the global motion case.
40 It has been proposed that cortical neurons are hyperexcitable in migraine, either due to hyperexcitability itself or secondary to reduced cortical inhibition.
1 7 37 41 42 43 Irrespective of the mechanism of hyperexcitability, it may be predicted to increase the level of neural noise and hence impair the ability to integrate local cues into global percepts when embedded in noise. In contrast, vernier thresholds are quite robust to noise presented in the form of positional jitter of elements.
44 If we assume that the positional noise for each dot comprising the vernier bars is independent and distributed in a Gaussian fashion, the effect of such noise on the average position of the overall bar will be minimal, if any. In contrast, for the global stimuli, variations in positional noise may change the local orientation of dot pairs or trajectory of movement. Consequently, if neural hyperexcitability results in increased noise in the form of positional uncertainty, it may be expected to impair performance on the global tasks but to have a lesser effect on the dotted vernier acuity tasks used in this study. Positional noise of this fashion would also be expected to affect minimally the orientation discrimination thresholds performed with sinusoidal grating stimuli and so is consistent with previous findings of normal performance of migraine groups for such tasks.
8
There are several other differences in the stimulus characteristics that may contribute to the finding of selective loss for the global tasks. One such difference is that the size of the dot elements comprising the vernier stimuli (17 sec arc) was much smaller than those comprising the global stimuli (4.5 min arc). The width of the vernier bars was 10 min arc which was similar to the spacing between the dot pairs in the glass patterns (9 min arc). The parameters for the vernier stimuli were chosen with the objective of maximizing the sensitivity for the motion-defined vernier stimulus,
25 whereas the stimulus parameters for our global tasks were chosen to be consistent with our previous research.
2 4 It has been demonstrated that individuals with unilateral headache are more likely to report grating-induced visual illusions than control subjects, and that this effect is most prominent for gratings with spatial frequencies of 3 to 4 cyc/deg.
45 The mechanism for this discomfort is unclear; however, it has been proposed to be related to neural excitability.
45 Our vernier bars were 10 min arc (hence equivalent to a single bar of a grating of 3 cyc/deg). It is possible that migraineurs may have more difficulty with the vernier task if presented as a grating alignment task or for different bar widths or dot element sizes. These further investigations fall outside the scope of the present study.
Another possible contributing factor is stimulus duration. In this study, the vernier stimuli were presented for 456 ms and the global tasks for 400 ms. These durations were chosen to be consistent with our previous work.
2 4 Several previous studies have identified contrast-processing deficits in migraineurs when briefly presented stimuli were used (30 ms or less); however, these have been presented in the context of adaptation or masking experimental designs.
6 46 It is possible that patients with migraine make more errors in judgment with briefly presented vernier targets than do nonheadache control subjects. This issue also falls outside the scope of the present study.
Aberrant psychophysical thresholds in clinical populations may also arise at times, due to deficits in attention rather than perception.
47 Given that different migraineurs displayed the worst global motion and global form thresholds (see
Fig. 5 ), is it possible that the data reflects inattention? We have recently completed a study comparing neuropsychological measures in migraine with global motion perception and found no evidence for an attention deficit or other cognitive difference between migraine and control groups.
48 It is also not clear why inattention would exclusively influence performance on the global tasks and not the vernier tasks. The psychophysical methods were the same for all tasks, hence incorrect responses due to inattention should be equally damaging to the estimated thresholds. Consequently, it seems unlikely that the differences between groups can be explained by differences in attention. Instead, the different performance on the global motion and global form tasks within individuals could be explained by different levels of instantaneous noise during the tasks, or possibly by individual variations in the degree of hyperexcitability in different areas of the cortex.
In the present study, we found no significant difference in performance between the MO and MO groups. Hence, impaired performance on the global tasks was not a consequence of the aura process. This finding is consistent with those of other studies of visual processing in migraineurs with and without aura.
2 3 4 46 49 50 Measures using electrophysiology and TMS have also been shown to be similar in aura and nonaura groups.
15 51 Many individuals experience migraine events both with and without aura during their lifetimes; hence, these migraine conditions may form part of a continuous spectrum rather than being truly separate entities. Nevertheless, there are reports of more pronounced visual deficits in individuals with aura than without.
6 52 Further study with larger sample sizes is needed to determine conclusively whether more severe visual processing anomalies are present in groups with aura; however, both the present study and previous literature demonstrate that visual aura is not necessary for presence of interictal visual processing abnormalities. We also did not find any relationship between psychophysical thresholds and headache characteristics. If the global task performance is impaired due to aberrant neural processing that is inherent in the migraine process, then it may be expected that findings would not worsen in a cumulative fashion with increasing length of migraine history. Our results should be treated cautiously due to the relatively small number of subjects involved and the often unreliable nature of subjective retrospective reporting of migraine symptomatology.
In summary, the present study demonstrates that, as a group, migraineurs perform more poorly than do nonheadache control subjects on tasks assessing the integration of local motion or form cues into global percepts. The ability to make spatial localization judgments is not impaired. This finding does not preclude V1 involvement but demonstrates that deficits on performance of extrastriate tasks do not arise due to dropout of information at V1. It seems likely that aspects of stimulus design are critical for identifying deficits of visual processing in migraine. Aberrant cortical neural processing is not isolated to either form or motion processing and, taking into consideration the results of both this study and previous works, is likely to involve both striate and extrastriate cortical visual areas. The possible functional significance of such deficits for individuals with migraine requires further study.