Conventional Pattern-Reversal VEPs Are Not Equivalent to Summed Multifocal VEPs

Abstract

purpose. To compare conventional pattern-reversal visual evoked potentials (cVEPs) with multifocal VEPs (mfVEPs).

methods. mfVEPs and cVEPs were recorded during the same session in 12 normal subjects with an active electrode at Oz referenced to the inion (Oz-In) or to a midfrontal position, Fz (Oz-Fz). The mfVEP stimulus, a 60-sector dartboard, had a mean luminance of 100 cd/m² and a diameter of 42.2°. The cVEP checkerboard stimulus subtended 21°, had a mean luminance of 75 cd/m² and a contrast of 90%. Transient responses (2.5 Hz) were recorded for check sizes ranging from 12 to 50 minutes of arc (minarc). White cardboard masks were used to isolate upper and lower hemifields, within various field windows, for comparison with corresponding parts of the mfVEP. In a second experiment, VEPs were obtained using slowed m-sequences (8 and 16 video frames per m-step), as well as square-wave periodic reversals (2.4 Hz), for both the scaled dartboard display and an unscaled checkerboard display (check size of 50 minarc).

results. The mfVEPs to fast m-sequence stimulation showed a strong polarity reversal between waveforms from the upper versus the lower hemifield. The cVEPs had larger amplitudes (∼3×) and longer implicit times (∼15–20 ms) and did not show the polarity reversal. Amplitude asymmetry between upper and lower hemifields was larger for cVEPs than for mfVEPs. As the stimulation rate was slowed, response amplitudes and implicit times of the major features increased, the upper versus lower polarity reversal was generally lost, and asymmetry of hemifield amplitudes grew. The same pattern of results was observed for scaled and unscaled spatial displays and for Oz-Fz and Oz-In signal derivations.

conclusions. Full-field cVEPs cannot be simply related to the sum of mfVEPs when each are recorded under their typical conditions. The stimulation rate has the largest influence on the differences between the two response types. The findings suggest that contributions from extrastriate sources are greater with the cVEP paradigm or the slowed mfVEP sequence than with the standard mfVEP paradigm.

The visual evoked potential (VEP) has long been used for evaluation of function in glaucoma. ^{1 2 3 4 5 6 7 8 9} Traditional VEP techniques, ^{10 11} however, as applied to the study of glaucoma, are limited by relatively low spatial resolution, as well as by the dominance of contributions from foveal and inferior visual field areas. ^{1 3 12 13 14 15 16 17 18 19} In fact, the relative contribution from different visual field areas to the large-field VEP also depends on the unique anatomy of each subject, as well as the particular stimulus characteristics and electrode positions used. ^{14 17 20 21 22 23} Given that the anatomy of the striate cortex (primary visual cortex, V1) can vary greatly among individuals, ^{24 25 26 27} and that extrastriate areas can also contribute to VEP responses, ^{13 17 18 20 21 28 29 30 31 32 33 34 35 36} it is not surprising that a wide range of normal limits often constrains clinical VEP studies. 

The multifocal technique has improved the clinical utility of VEP testing by enabling much finer spatial resolution without a significant increase in recording time. ^{37 38} Analysis of individual field locations reduces some of the confounders inherent in the traditional large-field VEP. Moreover, the typical multifocal VEP (mfVEP) display is spatially scaled with eccentricity, which serves to increase the extent of the visual field that may be assessed simultaneously. ^{16 37 39} Given these improvements, the mfVEP has recently received much attention regarding its application to glaucoma and other optic nerve diseases. ^{40 41 42 43 44 45 46 47 48 49}  

Yet, to date, little work has been done directly comparing mfVEPs with conventional VEPs. Barber ⁵⁰ and Barber and Wen ⁵¹ compared mfVEP responses to fast m-sequence modulation with conventional (large field) pattern-onset responses in a few subjects. In general, there were gross similarities between the sum of the mfVEPs and the large-field pattern-onset VEP, although there were also salient differences. Thus, our understanding of the relationship between mfVEPs and conventional VEPs is incomplete. Furthermore, Barber did not report on comparisons between mfVEPs and conventional pattern-reversal VEPs, perhaps the more commonly used clinical stimulus. Hence, the relationship between these two methods remains unknown. 

The purpose of this study was to compare the relationship between the mfVEP and the conventional transient pattern-reversal VEP (hereinafter, cVEP). We find that the mfVEP records for the most typically used stimulus (fast m-sequence) are not equivalent to the individual parts of the cVEP. 

Methods

Experiment 1

Subjects.

Twelve healthy, volunteers with normal vision, seven men and five women, participated in the study. The mean age (±SD) of the group was 35.6 ± 10.5 years. Five of these subjects completed an additional testing session, with a more extensive protocol (see experiment 2). All subjects met the following inclusion criteria: good visual acuity (≥20/20), intraocular pressure of 22 mm Hg or less, symmetrical optic discs (asymmetry of vertical cup to disc <0.2) normal standard achromatic increment threshold visual fields (Swedish interactive test algorithm [SITA] 30-2 or 24-2 results within normal 95% confidence limits for both mean defect [MD] and pattern standard deviation [PSD]; Humphrey Systems, Dublin, CA) and glaucoma hemifield test (GHT; Humphrey Systems) results within normal limits. None of the subjects had a history of ocular disease, surgery, trauma, or systemic disease known to affect the visual system. 

All subjects were fully informed with regard to the potential risks and benefits of the study and then provided voluntary written consent to participate. All procedures complied with the tenets of the Declaration of Helsinki and were preapproved by the Legacy Health System’s institutional review board for the protection of human subjects. 

Multifocal VEPs.

Multifocal VEPs were recorded and analyzed using a visual evoked response imaging system (VERIS, ver. 4 software; Electro-Diagnostic Imaging, San Mateo, CA). Gold disc electroencephalogram electrodes were placed according to the International 10-20 system, with the active electrode at Oz (∼3.5 cm anterior to the inion, on average, for this group of subjects). According to conventional VEP standards, the active electrode was referenced to the midfrontal position (Fz, ∼12 cm posterior to the nasion) for one derivation (Oz-Fz). For the other derivation, one most commonly used for mfVEPs, the active electrode was referenced to the inion (Oz-In). The left earlobe served as ground. The skin was prepared with a gel on a cotton swab (Nuprep; D. O. Weaver & Co., Aurora, CO). Electrodes were fixed in position with conductive cream (EC2; Astro-Med, Inc., Warwick, RI) and secured with a wrap (Coban; 3M, St. Paul, MN). Electrode impedance was maintained below 5 kΩ in all cases and was usually below 2 kΩ. 

Signals were amplified by 100 k (Grass Model 12, Astro-Med, Inc.), band-pass filtered 3 to 100 Hz (1/2 amplitude) and sampled at 1200 Hz (0.83-msec bins). Stimuli were presented on a 21-inch monochrome monitor (Nortech, Plymouth, MN) with a 75 Hz refresh rate. The mfVEP stimulus (Dart Board 60 With Pattern; see Fig. 1A ) consisted of 60 sectors, each with 16 checks, 8 white (200 cd/m²) and 8 black (<1 cd/m²) providing a Michelson contrast of approximately 99%. The size of the individual stimulus areas, and correspondingly, their check sizes, are approximately scaled with eccentricity according to a cortical magnification factor. ^{37 38 39} The stimulus had a total diameter of 42.2° at the test distance of 35 cm. 

Contrast reversals were determined by a pseudorandom m-sequence with a total of 2¹⁵ − 1 steps that required approximately 8 minutes to complete. Each test session was divided into 16 segments, and subjects were allowed to rest between segments, as necessary. For experiment 1, each subject completed two sessions per eye in an ABBA-order, with the first eye chosen at random. For experiment 2, only the right eye was tested. 

All stimuli were viewed through natural pupils with optimal refractive correction in place. (For a more detailed description of the general multifocal technique, see Refs. ^{52 53 54} .) 

Conventional VEPs.

Conventional pattern-reversal VEPs were recorded during the same session (Utas-E3000 system; LKC Technologies, Gaithersburg, MD) in a manner adhering to recommended international standards. ^{10 11} Checkerboard stimuli were presented on a monitor subtending 28° × 21° at 57 cm, with a mean luminance of 75 cd/m² and contrast of 90% (see Figs. 1C 1D 1E ). Transient pattern-reversal responses (five reversals per second) were recorded to patterns with check sizes of 12, 25, and 50 minarc. White cardboard masks were used to create upper hemifield, lower hemifield, and full-field stimuli for three displays: a central 5° window (2.5° radius; Fig 1C ), a 20° window (10° radius; Fig 1D ), and a 5° to 20° annular window (2.5 and 10.0° inner and outer radii; Fig 1E ). Checks of 12 and 25 minarc were used with the central 5° field window, and checks of 25 and 50 minarc were used with the annular and full 20° windows. 

These field windows were chosen to facilitate comparisons with the analogous regions of the mfVEP dartboard stimulus (Fig. 1A , inset). In particular, the central 5° window for cVEPs had the same area as the group of 12 mfVEP stimulus sectors within the central two eccentricity groups. Similarly, the 24 mfVEP sectors within the 5° to 20° annulus (two surrounding eccentricity rings) had the same area as the annular cVEP condition. As mentioned, two check sizes were chosen for each of the three cVEP field conditions, so that they spanned the range of check sizes within the corresponding region of the scaled mfVEP stimulus (dartboard). 

Figures 1C 1D 1E illustrate nine of the conditions. Figure 1C shows the central 5° field window (check size of 25 minarc), Figure 1D shows the 20° window, and Fig. 1E shows the 5° to 20° annular window (check size of 50 minarc in D and E). The left column in Figures 1C 1D 1E shows the full-field condition, the middle and right columns show the upper and lower hemifield conditions, respectively. For eight subjects, the stimulus with the 50-minute checks was omitted during the 20° window condition because the first four subjects had all produced larger responses to the 25-minute checks in their test sessions. Room lighting was diffused and set so that the luminance of the cardboard masks was close to the mean of the checkerboard stimulus. 

Signals were amplified by 40 k and sampled at 2000 Hz from two channels (Oz-Fz and Oz-In). High- and low-pass filters were set at 1 and 100 Hz, respectively, for both channels. Two records were obtained for each stimulus condition, for each eye. Each record was an average of either 100 or 200 sweeps, depending on the signal-to-noise characteristics, obtained in groups of approximately 25 sweeps with brief pauses between each group. The length of each record was 256 ms. Automatic artifact rejection was set to ±75 μV (60% of the analog-to-digital range). All subjects were highly trained and practiced observers and were instructed to maintain accurate fixation during all recordings. 

Data Analysis

The results of two mfVEP sessions for each eye were averaged. Spatial smoothing and artifact rejection features available with the VERIS software (Electro-Diagnostic Imaging) were not used in any of the analyses. The principal analyses were performed on the first slice of the second-order kernel, which represents the mean difference between responses to frame pairs when a reversal occurred versus frame pairs when no reversal occurred. Thus, it represents the response kernel for local contrast reversals between successive video frames and is therefore analogous to the conventional pattern-reversal response. Recent versions of the software flip the polarity of the second-order kernel (relative to the first-order kernel; e.g., compare Fig. 2 in Ref. ⁵⁴ , with Fig. 1 in Ref. ⁵³ ). Hence, in this article, the mfVEP records are shown positive-upward with their polarity flipped back in relation to the software output. 

The mfVEP records were summed for various regions of the field that corresponded to the windows used for cVEPs (Fig. 1) , including the full window and the upper and lower hemifields, for the central 5°, the central 20°, and the 5° to 20° annular windows. mfVEP records from these regions were compared with the cVEP responses from similarly isolated stimulus regions for patterns with similar check sizes (described earlier). Response polarity, peak implicit times, and peak-to-trough amplitudes were measured, and the group means were compared. 

Results

Figure 2 compares the cVEPs (left column) and mfVEP records (middle and right columns) for one representative subject. The three pairs of rows show responses for the three field windows described in Figures 1C 1D 1E . In each row, the cVEP responses shown are for check sizes that are approximately equal to the average check size within the corresponding region of the scaled multifocal stimulus (dartboard). For example, the top pair of rows shows the central 5° cVEP condition using the 12-minute checks in the left column, and the center and right columns show the Oz-In and Oz-Fz mfVEPs, respectively, from the central 5° of the dartboard where the checks of the sectors average approximately 12 minutes. Responses for the upper and lower hemifields are shown in the top row for each field area. For the cVEP, these responses were obtained by occluding the display (see the Methods section and Fig. 1 ), whereas for the mfVEP, the appropriate responses were summed. 

For cVEPs (standard Oz-Fz recordings, first column), the responses to isolated lower hemifield stimuli (solid black traces) were larger than those for equivalent stimuli presented to the upper hemifield (dashed black traces). The upper and lower hemifield responses also had the same polarity and similar timing for all three field conditions. (Note that small fixation drifts would have little impact on the asymmetry of hemifield response amplitude.) 

In contrast to the cVEP findings, the mfVEP records (middle and right columns) summed for the lower hemifield had polarity opposite to that of records summed from the upper hemifield, in both the Oz-In and Oz-Fz recordings (middle and right columns, respectively) for all three field areas evaluated (central 5°, central 20°, and 5°– 20° annulus). Similar to the cVEP, the lower hemifield records were larger than the upper for all three field areas, although the asymmetry was not as marked as it was for cVEPs. The cVEPs were considerably larger than the mfVEPs. Note that the scale was twice as large for the mfVEPs. Also note that the summed mfVEPs for the upper hemifield had polarity opposite to that of the upper hemifield cVEP responses, even for the same derivation (Oz-Fz). 

The dashed gray traces, in the second row for each field window, show the sum of upper and lower hemifield VEPs. For the mfVEP, the sum was always smaller than the parts, because the responses to the hemifields had opposite polarities. For the cVEP, the sum of the upper and lower hemifield responses (dashed gray traces in the left column) was larger than the actual response to combined stimulation (solid gray traces) for all three field conditions shown for this subject. This will be explored in more detail later, for the entire subject group. 

Figure 3 shows the group mean (±SEM) implicit times for the first three features of the cVEPs (Fig. 3A) and mfVEPs (Fig. 3B) . The results are shown separately for right eyes (open symbols) and left eyes (closed symbols). Both panels show the data for Oz-Fz recordings. The traces shown are inferior hemifield responses, for one representative subject, to illustrate typical response features. As expected, the cVEP responses consistently showed a trough at ∼85 ms, a peak at ∼110 ms, and another deeper trough at ∼155 ms (similar to the features normally labeled N75, P100, and N135, respectively). As seen in the subject’s data shown in Figure 2 (middle left), the responses to checkerboard pattern reversal (25-minute checks) have similar timing for upper and lower hemifield, as well as full-field stimulation. Although not statistically significant, there was a tendency for the P100 of upper-field responses to be slightly (∼3–4 ms) slower than that for lower field responses. 

A negative-positive-negative waveform was also consistently observed for lower hemifield mfVEP traces, however, as seen for the subject in Figure 2 (middle right) the polarity of the features was opposite for upper hemifield traces. Thus, the three successive peaks (or troughs) cannot be identified using the common polarity designations P or N, and so only the numbers 1, 2, and 3 are used. 

Figure 3B shows that the implicit times for mfVEP records were consistently approximately 15 to 20 ms faster than the cVEPs for the corresponding hemifield. There was slightly larger variability for the implicit time of the third mfVEP feature compared with the N135 of the cVEP. Upper hemifield mfVEP implicit times were also longer than lower hemifield responses, on average, for the group (P < 0.01 ANOVA). 

Figure 4 summarizes the group amplitude data for the same conditions reported in Figure 3 (20° field window). In Figure 4A , right, are the records for full-field (gray), upper (dashed), and lower hemifield stimuli (solid black traces) for one subject. The group means (±SEM) for the two peak-to-trough amplitude measurements are represented graphically at left. The results are shown separately for the right eyes (open symbols) and left eyes (closed symbols). To facilitate cross-comparisons with mfVEPs throughout the remainder of this report, the labeling scheme reflects the order of appearance of each response feature (i.e., the implicit time reference was dropped). As was seen for the single subject in Figure 2 , the group data show that the lower hemifield cVEP responses were more than twice as large as the upper hemifield responses and had the same polarity. The upper and lower hemifield cVEPs had the same polarity for all 12 subjects under this stimulus condition (25-minute checks, whole-field window). Two of the 12 subjects showed a tendency toward polarity reversal between the upper and lower hemifield cVEP responses recorded in the Oz-In channel. The group data for Oz-In cVEPs, however, were very similar to the Oz-Fz data, only approximately 50% smaller and slightly more variable (data not shown). 

The group average cVEP data (Fig. 4A) also show that the sum of the upper and lower hemifield responses was larger than the amplitude of the full-field response. That is, if the hemifield response amplitudes were to sum linearly, the p1-n1 and p1-n2 amplitudes would be ∼12 and ∼19.5 μV, compared with the observed full-field response amplitudes of ∼9 and ∼14 μV, respectively. This phenomenon was observed for 11 of the 12 subjects when this stimulus condition was used (there was no difference for the 12th subject). 

In Figure 4B , right, are the mfVEP records (Oz-Fz) for the same subject shown at the right in Figure 4A , summed for the upper (dashed black trace) and lower (solid black trace) hemifields, as well as for the full 20° window of the dartboard stimulus (dashed gray trace). The group mean (±SEM) data are shown in the graph at the left for the Oz-Fz derivation (although not typically used for mfVEP, the Oz-Fz data are shown for direct comparison with the cVEP responses). The sum for the whole is, by definition, equal to the sum of the parts of the mfVEP, thus the full-field data are shown only for comparison to the cVEP data (rather than with the sum of the parts, as was done above for cVEPs). The group average data clearly show that the polarity of upper and lower hemifield records were opposite to each other. The polarity reversal between mfVEPs summed for upper and lower hemifields was observed for all 12 individual subjects. The group average data also show that the polarity of upper hemifield cVEPs was opposite that for mfVEPs (compare Figs. 4A with 4B and 4C ). 

Similar to that seen for cVEPs, asymmetry in hemifield amplitude was also observed in mfVEPs with the Oz-Fz derivation. The amplitude of lower field sums were approximately 2.5 times larger, on average, than that of upper field sums. The dominance of the lower hemifield can also be seen for mfVEPs by inspection of the size and polarity of the full-field summed response in Figure 4B . If the hemifields produced identical responses of opposite polarity, the sum would be a flat line (with zero amplitude). 

Figure 4C shows the mfVEP data for the Oz-In derivation, the typical recording mode for mfVEPs. At right are the waveforms for the individual subject and at left are the group mean data (±SEM). For the Oz-In derivation, there is less asymmetry between upper and lower hemifield response sums. The group data show that the lower hemifield sums were only approximately 1.12 times larger, on average, than the upper hemifield sums. The group mean for the mfVEP full-field sum is closer to zero, although still slightly dominated by the lower field response polarity. Again, all 12 individual subjects showed a strong polarity reversal for mfVEPs summed within the upper versus the lower hemifield. 

Table 1 shows the group average values (±SD) for implicit times and amplitudes for cVEPs (25-minute checks, full-field window) and mfVEPs of both Oz-Fz and Oz-In derivations. It is clear that there are important differences between VEP responses elicited by the conventional large-field, transient pattern-reversal method compared with the multifocal method. The differences include the dramatically opposite relationship between the polarities of upper versus lower hemifield records, as well as a modest difference in response timing. Some of the differences found for implicit times might be explained by the fact that the luminance and contrast of the cVEP stimulus were both slightly lower than the mfVEP stimulus. However, the implicit time differences were larger than would be expected given the relatively small differences in mean luminance (∼0.125 log units) and contrast (∼10%). Nevertheless, in a second experiment an attempt was made to compare the two response types directly using the same stimulus monitor, mean luminance, and contrast. 

Experiment 2: Effects of Temporal Rate and Mode of Stimulus Modulation

Methods.

Five subjects completed a separate recording session during which the stimulus conditions were systematically varied between those typically used for mfVEPs and those similar to the cVEP stimuli as described earlier. For this experiment, all stimuli were generated, and all records were obtained, using VERIS (Electro-Diagnostic Imaging). All stimuli were presented with a mean luminance of 100 cd/m² and Michelson contrast of 99% on a monitor with a refresh rate of 75 Hz (similar to the standard mfVEP stimulus described earlier). All signals were amplified by 100 k, band-pass filtered 3 to 100 Hz (one-half amplitude) and sampled at 1200 Hz (0.83-ms bins) using the same two channel derivations, Oz-Fz and Oz-In. There were four temporal and two spatial stimulus conditions, with three field-masking conditions for each combination. The four temporal conditions included: (1) the standard fast m-sequence, in which the duration of each m-step is equal to one video frame (designated “std-m”), (2) a moderately slow m-sequence that has eight frames per m-step (8m), (3) a slower m-sequence with 16 frames per m-step (16m), and (4) a square-wave periodic reversal with a rate of 4.7 reversals per second (a reversal every 16th video frame). 

Two spatial arrangements were used with each of the above temporal sequences: The first was the scaled “dartboard 60,” as described earlier (Fig. 1A) , and the second was the unscaled checkerboard (Fig. 1B ; a custom stimulus available with VERIS, ver. 4.3, electro-Diagnostics). For the latter, the check sides subtended 50 minutes at the 35-cm test distance, which is close to the average check size in the fourth ring of the scaled dartboard stimulus, as well as the largest check size used for cVEP stimulation in experiment 1 (i.e., with the LKC system). The total field of the unscaled checkerboard was 39° × 39°. 

The length of recording for each test run depended on the temporal sequence. The standard m-sequence with dartboard stimulus had an exponent of 15 (as shown earlier) and required 7 minutes 17 seconds to complete. The other m-sequence tests runs all had an exponent of 12, and thus the same number of stimulus presentations. Consequently, the std-m condition for the unscaled checkerboard (which acts as a single element) required only 55 seconds to complete, whereas the 8m condition required 7 minutes 17 seconds, and the 16m condition required 14 minutes 33 seconds. The square-wave reversal runs required 1 minute 49 seconds to complete (the exponent was 8, providing 255 steps, each 213.33 ms). 

For the m-sequence tests using the dartboard stimulus, records could be summed for any spatial region of interest, as is normally done using the VERIS software (Electro-Diagnostic Imaging). As in experiment 1, the responses from upper and lower hemifields were summed. However, for the other stimuli, white cardboard masks were needed to isolate the hemifields of interest, because the entire field acts as one element. Thus, one test run was completed for each of three field window conditions, the full-field (i.e., no mask), an upper hemifield only, and a lower hemifield only. 

Results.

Figure 5 shows the full set of results for one subject. The grid is arranged from top to bottom by temporal rate (modulation sequence) and from left to right by spatial pattern (scaled dartboard left two columns, unscaled checkerboard right two columns) and signal derivation (first and third column, Oz-In; second and fourth columns, Oz-Fz). Thus, at the top left are the records for the typical mfVEP method and at the bottom right are the records for the standard cVEP method. As one moves down the rows, the stimulation rate changes from the fast (standard) m-sequence, to a slowed m-sequence (8m, row 2), to an even slower m-sequence (16m, row 3), and finally to the square-wave periodic reversal stimulus (bottom row). As in previous figures, records for the lower hemifield are shown by thin, solid black traces, upper hemifield by bold dashed black traces, and full-field by light gray traces. Light gray traces are omitted in the three m-sequence-with-dartboard conditions because, by definition, the full-field record equals the sum of the upper and lower hemifield records. For the other conditions, the full-field responses were acquired independently from the hemifield responses (see the Methods section, Experiment 2). 

There were three obvious changes that occurred as the method was altered from standard mfVEP to standard cVEP (i.e., top left to bottom right). First, a strong asymmetry developed between the size of the lower and upper hemifield records. Marked hemifield asymmetry was present with all but the standard m-sequence. Second, the implicit times of the major features increased. Third, the polarity reversal between the upper and lower hemifield responses was lost. The temporal sequence (down the rows) has a stronger influence on these three changes than did the spatial pattern or recording configuration (across the columns). For example, all the standard m-sequence conditions (first row) showed a polarity reversal, whereas little or no polarity reversal was seen for the slowest sequences in rows 3 and 4. 

As an aid, + appears below each pair of hemifield records showing polarity reversal, – when the pair do not show polarity reversal, and ? if the reversal is ambiguous. These judgments about polarity reversal were based on subjective analysis of first and second major features of the waveform. Although individuals may have differed about a particular set of responses, the trend was clear. Dramatic amplitude asymmetry developed and polarity reversal was lost as the m-sequence was slowed from the standard fast condition to the slow (8m) condition. This pattern of change occurred regardless of signal derivation (Fig. 5 ; compare columns 1 with 2 and 3 with 4). The results for the scaled dartboard stimulus followed the same general pattern for amplitude asymmetry as the unscaled display, but the loss of polarity reversal was less clear for some of the waveform features. In particular, the records for the upper hemifield were more complex than the lower hemifield records (Fig. 5 ; see especially lower left), and the loss of polarity reversal was clear only for some of the early features (before ∼100 ms). This will be evaluated further later (Fig. 8) . 

A corollary to the results described earlier is that the full-field responses became dominated by the lower hemifield as the stimulus became more similar to that used for cVEP. The right column in Figure 5 clearly shows that this effect was due mostly to temporal rate. The amplitude of the full-field response grew dramatically as the sequence was slowed, which was due to several factors, including loss of hemifield polarity inversion, increased dominance of the lower hemifield, and overall growth in amplitude. 

Figure 6 shows the results in three other subjects to the scaled dartboard stimulus. There are two columns of records for each subject, the Oz-In channel derivation on the left (the same as column 1 in Fig. 5 ) and the Oz-Fz channel derivation on the right (the same as column 2 in Fig. 5 ). The same general results observed for subject 1 (Fig. 5) were recorded in these three subjects. First, the polarity reversal was lost as the temporal rate was slowed. This observation was most salient for the earliest part of the records and for the Oz-In signal derivation. Second, the implicit times of the major features increased. Third, the amplitude asymmetry, whereby lower hemifield records are larger than upper hemifield records, grew as the temporal rate was slowed (also see Fig. 7 ), similar to that for subject 1, the amplitude grew dramatically as the stimulus was slowed. The full-field responses were also dominated by the lower hemifield for the periodic reversal stimuli (Fig. 6 , compare solid traces in row 4 with row 5). 

Figure 7 graphs the mean (of five subjects ± SEM) lower-to-upper hemifield amplitude ratio for five of the eight conditions along with the cVEP results from experiment 1 (LKC). Hatched bars show the Oz-In data and filled bars the Oz-Fz data. Amplitude was measured by the caliper method for the feature closest to 100 ms (e.g., the amplitude of the P100 for cVEPs would equal the voltage difference between the value at P100 and the line connecting the N75 to the N135). For the standard m-sequence (left pair of bars in Fig. 7 ), on average, the amplitude of the lower hemifield record was approximately 1.5 times larger than the upper hemifield and had the opposite polarity (i.e., negative ratio value). As the m-sequence was slowed to 8 frames per m-step (m8) and further slowed to 16 frames per m-step (m16), the polarity reversal was lost (positive ratio value) and the lower hemifield amplitudes become two to three times larger than the upper hemifield amplitudes. The same pattern of results was also seen for square-wave periodic reversal (SqWv-PR) using either the scaled dartboard (DB) pattern or the unscaled checkerboard (CB) pattern, as well as for the unscaled checkerboard pattern used by the system (LKC) for the cVEP stimulus in experiment 1. 

Local Waveform Differences and Hemifield Asymmetries

The results for subject 1 presented in Figure 5 showed clear loss of polarity reversal for the unscaled checkerboard pattern as the m-sequence was slowed, but a more complex pattern of change for the scaled dartboard stimulus, due mostly to the apparent complexity of the upper hemifield waveform. These findings are explored further in Figure 8 . The dartboard stimulus results for subject 1 have been analyzed with finer spatial resolution by summing waveforms from three radial wedge-shaped areas in each hemifield in Figures 8A 8B 8C . Figure 8D shows the results with periodic contrast reversals (2.5 Hz) of an unscaled checkerboard display (25-minute checks) using a cardboard mask with a wedge-shaped window that was rotated to record each of the six individual segments (recorded using the LKC instrument, 20° window). In each panel, the black traces show the records from the vertical midline in the upper and lower hemifields, and the light and dark solid gray traces show the records from the right (temporal) and left (nasal) wedges, respectively, bordering the horizontal midline. The dashed gray traces show the sum for the hemifield in each case. 

Figure 8 shows that the records from the areas abutting the horizontal midline within each hemifield were all quite similar (compare light and dark gray traces in each of the eight rows). However, the records summed along the vertical midline were quite different from those along the horizontal midline (compare black to light and dark gray traces in each row). In the lower hemifield, the vertical midline records had faster peak times and larger amplitudes than the records from areas along the horizontal. This characterization also applies reasonably well to the upper hemifield for the fast m-sequence (top row), but the differences were slightly larger than those in the lower hemifield. The differences between sectors in the upper hemifield became more striking as the m-sequence was slowed. Under slower conditions, the vertical midline traces had nearly opposite polarity from those in the horizontal groupings. This pattern was observed in all five subjects in experiment 2. 

This observation explains much of the amplitude asymmetry that developed between upper and lower hemifields as the stimulation rate was slowed. The increasing difference in the waveforms within the upper hemifield resulted in cancellation when the upper hemifield sectors were summed. This was in contrast to the lower hemifield, where the waveforms were still similar enough to sum constructively. 

Another possible explanation for the amplitude asymmetry was that there may be a differential change in amplitude for different field areas as the temporal rate of stimulation is slowed. The peak-to-trough amplitudes for records summed within these six wedge-shaped areas all grew by approximately a factor of two, on average (n = 5 subjects), as the sequence was slowed from the fast m-sequence to the 8m and 16m conditions. There were no statistically significant differences between any of the six wedge-shaped area sums for either the ratio of 8m:m or 16m:m amplitude. Nor was there a significant difference between the 8m and 16m conditions for any of the six areas. Thus, most of the effect of temporal rate on the amplitude asymmetry of hemifield summed records was due to the relative cancellation in the upper hemifield as shown in Fig 8 , although a larger subject group may help to uncover significant topographic amplitude changes. 

Higher-Order Kernels

Of course slowing the m-sequence separates the nonlinear interactions (higher-order kernels) further in time, so that their effects on the shape of the first slice of the second-order kernel (2k-1) become limited to later epochs. That is, because the next reversal occurred at 106.7 ms in the 8m condition and at 213.3 ms in the 16m condition, the earliest possible effects of nonlinear interactions on the 2k-1 were limited to epochs after these time points. ⁵³ Therefore, as the stimulus was slowed, waveform changes would be expected for the 2k-1 if there were higher-order kernels present in the fast m-sequence responses. Additional changes in the 2k-1 are to be expected if the higher-order kernels themselves change. That is, in addition to the predictable shift in time of their effects, changes in the shape of the higher-order kernel slices are also manifest in the 2k-1. 

Figure 9 shows the two higher-order kernel slices that have a consistent shape and signal-to-noise amplitude for fast m-sequence mfVEPs. The left column shows the second slice of the second-order kernel (2k-2) and the right column shows the first slice of the fourth-order kernel (4k-1). The top row shows the records for the fast m-sequence, middle row 8m, and bottom row 16m. Solid traces are the lower hemifield sum and dashed traces the upper hemifield sum in all cases. In the fast m-sequence condition, the two higher-order kernels had opposite polarity, and the lower hemifield sums were approximately 50% larger than the upper hemifield sums. Thus, the hemifield summed records bear a relationship to each other similar to that observed for the 2k-1. As the m-sequence was slowed, the shape and timing of the 2k-2 changed, and the sum for the lower hemifield became much larger than that for the upper hemifield. The signal-to-noise ratio of the 4k-1 diminished as may be expected for such long-range temporal interactions. Thus, as the m-sequence slowed some of the changes that occurred in the waveform shape and field topography of the 2k-1 were due to changes in the higher-order kernel shape and topographic distribution, in addition to the predictable changes of their manifestations within the 2k-1. 

Discussion

Conventional transient pattern-reversal VEP responses to relatively large-field stimuli (herein called cVEPs) cannot be related simply to the sum of local mfVEP responses as they are typically recorded (i.e., with fast m-sequence stimulation). There were four clear differences between the cVEPs and the mfVEPs. First, mfVEPs to fast m-sequence stimulation showed a strong polarity reversal between waveforms from the upper versus the lower hemifield, as previously reported, ^{37 40 42} whereas the cVEPs did not (Figs. 2 4) . In fact, the polarity of cVEP responses for the upper hemifield was opposite that of summed mfVEPs for the upper hemifield. Second, the implicit times for the first three major waveform features were all faster for the mfVEPs than for the cVEPs (Fig. 3 , Table 1 ), even when compared directly using identical spatial patterns, contrast, and mean luminance (Figs. 5 6) . Third, the relative imbalance between upper and lower hemifield amplitudes was much larger for cVEPs than for mfVEPs. Last, the amplitudes of the cVEPs were much larger than the mfVEPs for all field windows evaluated (Figs. 2 4 5 6) . Notably, the same results were observed regardless of whether a conventional monopolar (Oz-Fz) or a bipolar (Oz-In) signal derivation was used, although the Oz-In recordings had better hemifield symmetry for fast-m conditions (Figs. 2 and 4 5 6 7 8 , Table 1 ). Taken together, these results indicate that the two response types cannot be related directly to each other. Furthermore, they suggest that the underlying sources of cVEPs and mfVEPs are dissimilar. 

Some of these findings are in agreement, whereas others contrast with the results of a previous study that compared cVEPs with mfVEPs. In particular, Barber ⁵⁰ and Barber and Wen ⁵¹ showed that conventional (i.e., sequentially averaged) VEPs and mfVEPs for pattern-onset stimulation were similar, in that they both showed a strong amplitude asymmetry (lower larger than upper) and polarity reversal across the horizontal midline. Barber concluded that the resynthesized mfVEP responses (kernel sums) were comparable to cVEP responses for relatively small isolated areas near fixation. However, some obvious differences in the waveforms were noted, in particular, for the second peak at approximately 165 ms. Moreover, although not mentioned, the cVEP responses were approximately three times larger than the mfVEPs for the same central-field locations. The two response types appeared to have similar timing, which contrasts with the results of the present study. Thus, we agree that the general form of the cVEP can be reflected in the sum of mfVEPs for a given area, but disagree in that there are also significant differences between them (as detailed earlier). 

Some of the discrepancy between the findings of this study and the prior one may be due to the different stimulation paradigms compared (pattern-onset versus pattern-reversal). ⁵⁵ For example, there are well-known differences between pattern-onset, -offset, and reversal responses, and it is possible that reversal responses are dominated more by offset than onset components. ^{21 56 57} A full discussion of the differences between pattern-onset, -offset, and reversal response components, however, is beyond the scope of the present study, in which we evaluated only the transient pattern-reversal cVEP paradigm. 

In this study, some of the hemifield amplitude asymmetry for fast m-sequence stimulation was shown to be due to greater inhomogeneity of responses within the upper hemifield than in the lower hemifield (Fig. 8) . The difference in waveforms confirms the results of previous reports on mfVEPs for fast m-sequence stimulation. ^{44 58 59} The location and orientation of the calcarine sulcus relative to external landmarks (e.g., the inion) varies among individuals, and this also affects upper versus lower hemifield symmetry. ⁴² As the m-sequence was slowed in the present study, however, the inhomogeneity of responses within sectors from the upper hemifield actually increased and the subsequent cancellation resulted in an even larger asymmetry between hemifields (Fig. 8) . There is evidence suggesting that at least part of the well-known hemifield amplitude asymmetry for cVEPs is also due to a similar cancellation effect. ^{15 18 28} Yet, this degree of inhomogeneity between similarly sized and shaped sectors within the upper hemifield is not a uniform finding for cVEPs. ²¹ The differences between studies, especially among those with fewer subjects, probably reflects the known interindividual variability of cortical folding and presumed dipole orientations. ^{24 25 26} Additional intersubject variability may also arise from differences in hemifield activation patterns within the striate cortex. ²⁷  

Howe and Mitchell ³ commented on the significant limitations that hemifield asymmetries place on the potential use of cVEPs for functional assessment of glaucoma. Recent methodological and analytical developments may help to overcome some of these limitations in studies using cVEPs (see Refs. ^{19 60} ). These limitations notwithstanding, the mfVEP appears to be a rather effective method for objective assessment of localized functional losses in glaucoma. ^{40 42 44 45 46 48 49} In a study that directly compared cVEPs and mfVEPs in patients with glaucoma, the mfVEP was shown to be vastly superior to the cVEP for detection of advanced damage in the upper hemifield. ⁶¹  

The present study showed that the temporal rate of stimulation was the most important factor determining the differences between cVEPs and mfVEPs. The different spatial displays and signal derivations typically used for each technique were less influential factors. As alluded to earlier, when the temporal rate was slowed, the characteristics of mfVEPs became more like cVEPs. The polarity reversal between upper and lower hemifields was lost, the amplitude asymmetry became larger (i.e., the field topography changed), response amplitude increased, and the implicit times of the three prominent early features became longer. In some subjects, loss of polarity reversal for hemifield sums was less clear, which may be due to different effects in sectors along the horizontal versus vertical midlines, particularly in the upper field (e.g., see Figs. 5 and 8 , subject 1). 

There were only negligible differences between responses to the slowed m-sequence stimuli (8m and 16m) and responses to the periodic-reversal stimulus (Figs. 4 5 6) . This suggests that the larger amplitudes found for cVEPs (compared with fast m-sequence mfVEPs) is not due to harmonic resonance of generators at the 2.5-Hz stimulus rate used in the case of the former. Some of the amplitude growth for slower stimulation may be due to changes in the state of contrast adaptation. ^{62 63 64} Further, differences that arise with slower stimulation may be due to increased contribution of evoked response components to other stimulus attributes such as motion onset and offset. ²¹  

One parsimonious and reasonable explanation for the differences found between cVEPs and mfVEPs is that the balance of underlying generators is altered as the stimulation rate is slowed. The polarity reversal between upper and lower hemifields is often cited as the signature of dipole generators located within the calcarine fissure where the striate cortex (V1) largely resides. ³⁶ Previous work has shown that the likely generators of mfVEPs for fast m-sequence stimuli are primarily located in V1. ³⁸ Another study comparing functional brain imaging techniques also found that white noise stimuli preferentially activate area V1. ⁶⁵ The characteristics of the fast m-sequence mfVEPs recorded in the current study were also consistent with the characteristics of signals generated (primarily) in V1. Their polarity depended on field location and they had faster major feature latencies than slowed m-sequence mfVEPs and cVEPs. Their robust nonlinear components (higher-order kernels) probably represent the action of contrast adaptation mechanisms which have also been shown to affect neural responses in V1 (e.g., see Ref. ⁶⁶ ). The higher-order kernels for fast m-sequence stimulation also showed the strong polarity reversal across the horizontal meridian. 

In contrast, although some of the earliest components of conventional pattern-reversal and pattern-onset VEPs are also thought to arise in V1, the dominant later components are thought to represent mostly extrastriate sources. ^{18 20 30 32 33 36} Convergent conclusions have also been drawn using magnetoencephalographic (MEG) techniques. ^{67 68} As the m-sequence rate was slowed in the present study, responses became much more similar to cVEPs, with a dramatic increase in the amplitude of features that were essentially identical with the classic dominant features of cVEPs, the P100 and N135. These features, in particular, were slower (by ∼20 ms), generally lost the characteristic polarity reversal for V1 responses, and were dominated by lower hemifield responses along the vertical midline. ¹⁸ Taken together, the contrasting results between mfVEPs for fast m-sequence stimulation and cVEPs suggest that the former are dominated by contributions from V1, whereas the latter are more heavily influenced by extrastriate sources. Source localization studies, using either VEP with multielectrode arrays and/or MEG methods to model dipole positions, in conjunction with functional brain-imaging techniques such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET) may be able to address this hypothesis by comparing responses to these two stimuli directly. 

 Supported by The M. J. Murdock Charitable Trust, Vancouver, WA (BF), and National Eye Institute Grant EY02115 (DCH).

 Submitted for publication May 3, 2002; revised August 14, 2002; accepted August 22, 2002.

 Commercial relationships policy: N.

 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.

 

Figure 1.

 

(A) Multifocal VEP dartboard stimulus. Size of the 60 individual stimulus sectors and checks are scaled, with eccentricity. (B) The unscaled checkerboard display used in experiment 2. (C–E) The conventional VEP stimulus of the LKC system (LKC, Gaithersburg, MD) shown with field window and hemifield masks. (C) Central 5° field window (with check size of 25 minarc depicted); (D) the 20° window; (E) the 5° to 20° annular window (check size of 50 minarc depicted in D and E). (C–E) Left column: full-field condition; center and right columns: upper and lower hemifield conditions, respectively. Relative scale in all panels is approximately accurate.

Figure 2.

 

Conventional VEP responses (cVEPs, Oz-Fz recordings, left column) and multifocal mfVEPs (Oz-In and Oz-Fz recordings, middle and right columns, respectively) in one representative subject. Solid black traces: lower hemifield records; dashed black traces: upper hemifield records; solid gray traces: full-field responses; dashed gray traces: sum of upper and lower hemifield responses. Top: records for the isolated central 5° window (Fig 1C) ; middle: the 20° window (Fig. 1D) ; bottom: the 5° to 20° annular window (Fig. 1E) .

Figure 3.

 

Implicit times for cVEPs (A) and mfVEPs (B), both with data for Oz-Fz recordings (20° window). Lower hemifield records for one subject are shown above each graph to illustrate waveform features measured. Data points plot group mean (±SEM) of 12 subjects for right eyes (open symbols) and left eyes (filled symbols). Implicit times for cVEP responses to the 25-minute checks are shown for full-field (diamonds), isolated upper hemifield (triangles), and lower hemifield (circles) stimuli.

Figure 4.

 

(A) Peak-to-trough amplitudes for cVEPs (Oz-Fz; 25-minute checks). (B, C) mfVEP amplitudes for the Oz-Fz and Oz-In channel derivations, respectively. Data points plot group mean ±SEM (n = 12); (open symbols) right eyes, (closed symbols) left eyes, (diamonds) full-field, (triangles) upper hemifield, and (circles) lower hemifield. Left column: trough-to-peak amplitudes for the first pair of major response features (left) and the following pair (right; e.g., p1-n1 and p1-n2, respectively, for cVEPs). All data are for 20° field window. Right column: corresponding records for one representative subject.

Table 1.

 

Group Averages for cVEP and mfVEP

	Implicit Times (ms)					Amplitudes (μV)
		Feature 1	Feature 2	Feature 3	p1-n1	p1-n2
cVEP (Oz-Fz)
Full field stimulus	85.5 ± 4.4	110.6 ± 3.9	157.9 ± 17.2	8.8 ± 4.1	14.3 ± 7.7
Upper field stimulus	84.8 ± 7.1	114.1 ± 11.6	158.9 ± 21.1	3.9 ± 1.5	5.9 ± 2.6
Lower field stimulus	85.3 ± 8.0	111.4 ± 6.2	157.2 ± 14.9	8.0 ± 4.4	13.4 ± 6.3
mfVEP (Oz-Fz)
Full field sum	72.0 ± 5.5	93.7 ± 13.2	131.3 ± 16.0	1.3 ± 3.7	2.8 ± 4.1
Upper field sum	75.7 ± 7.9	103.8 ± 6.3	142.7 ± 19.5	−2.1 ± 1.7	−1.1 ± 1.3
Lower field sum	69.3 ± 5.3	90.7 ± 7.5	128.1 ± 17.6	3.1 ± 0.9	3.9 ± 1.5
cVEP (Oz-In)
Full field stimulus	87.2 ± 6.9	110.6 ± 5.4	157.6 ± 12.8	4.0 ± 2.6	6.3 ± 4.4
Upper field stimulus	80.0 ± 17.3	111.3 ± 16.3	150.3 ± 15.9	1.3 ± 3.5	1.4 ± 3.9
Lower field stimulus	79.9 ± 9.5	107.5 ± 7.1	155.8 ± 14.1	4.7 ± 2.2	8.4 ± 3.8
mfVEP (Oz-In)
Full field sum	74.4 ± 8.4	90.8 ± 9.3	124.1 ± 9.1	0.6 ± 1.7	0.8 ± 1.6
Upper field sum	72.7 ± 4.7	100.1 ± 6.2	134.6 ± 14.8	−2.6 ± 1.4	−2.0 ± 1.2
Lower field sum	70.9 ± 4.6	95.1 ± 8.4	136.5 ± 16.5	2.7 ± 1.1	2.4 ± 1.1

Figure 5.

 

Direct comparison of VEPs for standard fast m-sequence modulation (top row), slower m-sequence modulation (8m, second row; 16m, third row), and square-wave periodic reversals (bottom row) for subject 1 (experiment 2). The data for the Oz-In and Oz-Fz channels are shown for the scaled dartboard stimulus (left columns) and for the unscaled checkerboard stimulus (right columns; 50-minute checks). Bold dashed traces: upper hemifield; thin solid black traces: lower hemifield; light gray traces; full-field records. +, polarity reversal; −, no reversal.

Figure 6.

 

Results of experiment 2 for three other subjects. Only results for the scaled dartboard stimulus are shown. Details are the same as in Figure 5 .

Figure 7.

 

Lower-to-upper hemifield amplitude ratio for six stimulus conditions: standard fast m-sequence (std fast m), slowed m-sequences (8m and 16m), square-wave periodic reversals (Sq Wv PR) using the scaled dartboard stimulus (DB) or the unscaled checkerboard stimulus (CB; 50-minute checks) in VERIS (Electro-Diagnostic Imaging), or the cVEP stimulus (25-minute checks) with the LKC system (LKC Technologies, Gaithersburg, MD). Data are the mean ± SEM of results in five subjects for the (▨) Oz-in and ( ) Oz-Fz signal derivation.

Figure 8.

 

Intrahemifield waveform comparisons for (A) fast m-sequence stimulation, (B, C) slower m-sequence stimulation (8m and 16m, respectively), and (D) square-wave periodic reversal stimulation for subject 1. Solid black traces: records from the area along the vertical midline in each hemifield (see inset at right), light and dark solid gray traces: records from the right and left wedges, respectively, bordering the horizontal midline; dashed gray traces: sum for the hemifield in each case.

Figure 9.

 

Higher-order response kernels for subject 1; second-slice of the second-order kernel (2k-2, left column) and first-slice of the fourth-order kernel (4k-1, right column). Solid traces: lower hemifield sum; dashed traces: upper hemifield sum.

The authors thank Karin Novitsky for excellent assistance with data collection and Chris A. Johnson, Xian Zhang, and Bryan Winn for many helpful discussions.