Abnormal Waveform of the Human Pattern VEP: Contribution from Gamma Oscillatory Components

Abstract

purpose. Stimulus-related oscillatory activity above approximately 20 Hz (gamma band) is observed in the firing rate and membrane or field potentials of neurons in the visual cortex of cat and monkey. Cortical mass responses in the same frequency range are evoked in humans by contrast stimulation and proved (partly) independent of conventional VEPs. Visual evoked responses (pattern VEPs) with abnormal waveform (quasi-sinusoidal or with bifid wave P100) are reportedly common in diseases affecting the visual pathways (e.g., multiple sclerosis). Contributions of gamma activity to the distorted VEP waveform are possible if the mechanisms of generation are differentially affected by brain disorders. This hypothesis was investigated in patients with documented impairment of the visual system.

methods. VEPs and oscillatory responses to contrast stimulation (central 9° of visual field; 1.3 to 5.0 cyc/deg; 80% contrast; reversal at 2.13 Hz) were recorded in 765 patients referred for standard visual testing and frequency analyzed. Controls were 40 healthy volunteers. The stimulus conditions and recording procedures replicated previous normative studies on the human oscillatory response.

results. Conventional VEPs were replaced by oscillatory responses comparable to those of controls in the unfiltered recordings of 9.8% of examined patients (with postchiasmatic disorders in 59.9% of cases). Signal amplitude in the frequency interval of the VEPs (below approximately 19 Hz) was significantly lower in the frequency spectra of patients than of controls but did not differ in the frequency range above 20 Hz, reflecting the oscillatory response.

conclusions. The human gamma oscillatory response mediating in cortical visual information processing further proved independent of the VEPs. It can contribute to VEP waveform distortion, and its observation in substitution of VEPs should be regarded as an indication of impaired visual pathway function.

Increased latency of the cortical evoked response to contrast stimulation (pattern-reversal or pattern-onset VEPs) is an established indication of impaired function in brain diseases affecting the visual pathways. ^{1 2} Although difficult to characterize and seldom taken into account (see, for example, Collins et al., ³ Urbach et al., ⁴ Brecelj, ⁵ Sawaguchi and Ogawa ⁶ ), abnormal or distorted VEP waveform, particularly in the time window of wave P100, is also common. Pattern-VEPs with P100 breaking up into two positive waves (W-shaped or bifid P100s) or with superimposed quasi-sinusoidal sequences of negative/positive waves have been described in multiple sclerosis (with an estimated incidence up to 45%), migraine, vascular disease, and other neurologic diseases. ^{7 8 9 10 11 12 13 14 15} Averaged pattern-VEPs are the result of distinct components that eventually combine in time and space ^{15 16 17 18 19 20 21 22 23 24 25 26 27} ; accordingly, distorted VEPs are thought to reflect impaired or abnormally distributed activation of the contributing brain structures. ¹⁵  

An oscillatory mass response in the gamma range (approximately 20–45 Hz), with peak frequency centered around 25 to 30 Hz, contributes to and can be separated with negligible filter distortion from conventional human VEPs to transient (reversal or onset/offset) contrast stimulation. ^{28 29 30 31 32 33 34 35 36} This response is consistent with the gamma oscillations in spiking rate and membrane or field potentials that originate from the tonic excitation of networks of inhibitory cortical interneurons and are observed in multiunit recordings from neurons of the cat and monkey visual cortex. According to prevailing hypotheses, gamma activity originates in, and appears intrinsic to, structures with laminar organization such as the cortex, is enhanced during sensory information processing, and is modulated in part by oscillations in the synaptic input. It provides a frequency- and time-related coding system mediating (at some stage of visual information processing) in the synchronization of cortical neurons due to respond selectively to the stimulus physical properties. ^{36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54} In humans, the oscillatory gamma range response to contrast is recorded by electric and neuromagnetic methods, is peculiarly phase-locked to stimulus, has faster time dynamics and shorter latency than VEPs, and has different orientations of cortical sources. It is thought to reflect generating mechanisms partially independent of those of the VEPs. ^{28 29 30 31 32 33 34 35 36}  

Oscillatory responses at 20–40 Hz to contrast transient stimulation were reported in the absence of the conventional broadband VEPs in patients with brain damage involving the visual system. ^{14 32} Although anecdotal, this observation suggests pathophysiological conditions in which stimulus-related oscillations unaffected by pathologic brain conditions disorganizing the VEP response can account for abnormal VEP waveforms such as those described in neurologic diseases. However, this hypothesis was not considered when the gamma oscillatory response was not fully characterized in animals and humans. ^{7 8 9 10 11 12 13 14} The purpose of this study was to confirm previous observations,to estimate the incidence among patients of predominant oscillatory responses in unfiltered VEP recordings, and to compare these oscillations with the stimulus-related gamma responses of healthy controls. 

Patients and Methods

Patients (n = 765) referred over a 22-month period to the Neuro-ophthalmologic Unit of the Department of Motor Science for electrophysiological vision testing were considered irrespective of diagnosis, visual acuity, age, or any other factor. All patients underwent standard ophthalmologic and neurologic examinations and computerized visual perimetry, brain CT, or MRI when appropriate. Forty healthy volunteers recruited from among students and staff (23 men; age range, 18–45 years; mean age, 31.3 ± 6.2 years), without evidence or history of ocular, neurologic or systemic disorders, and with best-corrected visual acuity better than 20/18 served as controls; their screening VEPs and electroretinogram results were within normal limits for the laboratory standards. ^{55 56} Patients and controls were informed of the recording procedures, and the tenets of the Declaration of Helsinki concerning human experimentation were followed. The study was approved by the local committees. 

The protocol for electrophysiological visual testing was consistent with international guidelines ^{55 56} and replicated procedures used in previous studies in humans documenting an oscillatory mass response to contrast. ^{14 28 30 33 34 57} Visual stimuli were vertically oriented gratings with a sinusoidal profile (central 9° of visual field at 75-cm viewing distance; mean luminance, 30 cd · m⁻²; contrast, 80%) produced on a display (Tektronix 608; Portland, OR) by a digital generator (Venus; Neuroscientific Co., Farmingdale, NY) and reversed at 2.13 Hz. Each stimulus was presented for 29.9 seconds. A fixation point on the screen was provided, and the eye position was monitored. For all patients, stimulation was monocular at three spatial frequencies (1.3, 2.5, and 5 cyc/deg). Electrodes (dermal Ag/AgCl) were conventionally positioned 5 cm laterally to the inion and 5% of the inion-to-nasion distance above inion; the reference electrode was at Fpz and the ground at Cz. Amplifier (Physio-Amp; Francesco Marazza, Monza, Italy) bandpass and gain were 0.5 to 300 Hz and 50,000×. The digital system generating the stimuli also controlled electrophysiological data acquisition, which started synchronously with stimulus onset. The original signal wasprocessed off-line by discrete Fourier transform (DFT; final resolution; 1.03 Hz), and the amplitude spectrum was computed for each recording. 

In previous studies on volunteers, high-pass digital filters with cutoff at approximately 19 Hz proved adequate to separate the 20 to 45 Hz oscillatory response from the prominent low-frequency VEP, with minimal distortion despite the contiguous frequency intervals. DFT, Wavelets analysis, and second-order Bessel filters were equivalent in this respect. ^{14 28 30 33 34 57} In this study, the raw signal was processed by DFT and high-pass filtered with cutoff at 19 Hz (i.e., all signal components below 19 Hz were set to zero); the inverse DFT was then computed, and the average was recalculated on the filtered signal in 100 epochs free of artifacts. Eight to nine consecutive waves of the oscillatory response were identified and will be hereafter referred to as negative (n)0, positive (p)1, n1, p2, n2, p3, n3, p4, and n4 on polarity and order of appearance. Methods and normative data are described in detail elsewhere ¹⁴ (Fig. 1) . 

The raw signals were also processed to characterize the frequency organization and time dynamics of activities phase-locked or non-phase-locked to stimulus (see Narici et al., ²³ Sannita et al., ²⁸ Sannita et al., ⁵⁷ Salmelin and Hari ⁵⁸ ). For this purpose, the signal was analyzed by means of a bank of digital filters implemented through fast Fourier transform. The frequency range (0.5–50 Hz) was partitioned into 36 empirically predetermined frequency intervals (with 15-Hz width), overlapping in 1-Hz steps and centered at frequencies from 8.0 to 43 Hz. The raw and squared values of the filtered signal were averaged. The portion of signal activity phase-locked to stimuli (integrated rectified average [RIA]) and the total signal activity irrespective of phase-locking to stimulus (temporal spectral evolution [TSE]) were estimated across frequencies and over time in 1-second windows. For each frequency, RIA and TSE were computed as the poststimulus temporal evolution of the square root of the averaged signal power and of the square root of the raw signal power, respectively. RIA represents the average amplitude, at each frequency interval, of the signal having a constant time/phase relationship with the stimulus. TSE is an indicator of the averaged signal amplitude, at each frequency interval, irrespective of the time and phase relationship with stimulus; in this respect, it is analogous to other descriptors, such as the field-power indicators. The locking index (LOI), that is, the ratio of the phase-locked activity to the total (locked and unlocked) signal activity (TSE), was also computed to provide normalized estimates of phase-locked activity. TSE and RIA (after normalization to the maximum TSE or RIA value of each subject/recording) and the locking index were averaged across subjects. The approach has been applied by several research groups to characterize oscillatory signals evoked or induced by sensory inputs; its rationale and mathematical bases are described in detail elsewhere. ^{23 28 34 35 57 58}  

Results

Controls

Pattern-reversal VEPs in which waves N70, P100, and N145 could be identified unambiguously with latencies and amplitudes within the laboratory normal limits were recorded from all subjects under the stimulus conditions of the study. In all recordings, the signal amplitude clustered into two main peaks of the frequency spectrum centered in the low (<19 Hz) and approximately 20 to 45 Hz frequency intervals representative of the VEPs and oscillatory response, respectively (apparent separation between peaks at approximately 17–19 Hz; mean peak frequencies at 9 ± 2 Hz and 27.5 ± 3 Hz). The amplitude of the peak at approximately 20 to 45 Hz was smaller than that on low frequencies in all subjects. Oscillatory responses (i.e., quasi-sinusoidal sequences of 8–9 positive/negative waves) were separated from the VEPs by high-pass filtering the original signal with 19-Hz cutoff, with latencies of the first and last recognizable waves measuring 58.9 ± 7 ms and 180.5 ± 16 ms (see Figs. 1 4 5 ). 

The signal amplitude estimated irrespective of the signal-phase relationship with stimulus (TSE) increased in the frequency interval below approximately 19 Hz and in the time interval of the VEP response while homogeneously distributed throughout the response time window in the frequency range above approximately 19 to 20 Hz. Bursts of activity phase-locked to stimulus (RIA) were detected in the frequency intervals below 19 Hz and above 19–20 Hz with the cluster at 20–45 Hz accounting for the transient oscillatory response and anticipating in time that of the activity below 19 Hz (VEPs). Time dynamics were consistent among subjects and were emphasized when the LOI was computed (see Fig. 6 ). These findings replicate previous normative studies. ^{14 28 34 35 57}  

Patients

An oscillatory gamma response was superimposed on VEPs with reduced amplitude or occurred in the absence of recognizable conventional VEPs on at least one hemisphere in the unfiltered recordings of 9.8% patients. Unlike controls, the signal amplitude of these patients’ recordings was clustered at approximately 20 to 40 Hz (mean peak frequency, 24.7 ± 3.7 Hz), with only a smaller peak in the frequency interval below 19 Hz accounting, in healthy subjects, for the low-frequency VEP (Figs. 2 3 4 5) . These oscillations were not modified by high-pass filtering at the cutoff frequencies used to enhance the oscillatory responses in healthy subjects. Removing the predominant oscillatory response by bandpass filtering at 20 to 40 Hz resulted in the identification of either a recognizable VEP with reduced amplitude or no VEP response. Examples are given in Figures 2 and 3 . 

Statistical comparison versus controls indicates for the patient group a significant reduction of amplitude in the low-frequency interval representative of the conventional VEPs (0.5–19 Hz; t-test, t values between 5.56 and 11.08 for each frequency component; P < 0.0001); instead, no differences between patients and controls were observed in the frequency interval above 20 Hz, corresponding to the oscillatory response (Figs. 4 5) . A single burst of activity phase-locked to stimulus (RIA) was detected in the frequency interval above 19–20 Hz with time dynamics that were consistent among patients and that were emphasized when the LOI was computed; activity phase-locked to stimulus was not detected in the frequency interval below 19 Hz (VEP; Fig. 6 ). These observations were replicated in 14 patients referred after 3 to 5 months for follow-up. 

Postchiasmatic damage was documented in 59.9% of these patients, whereas prechiasmatic (retina or optic nerve) damage and pathologies either unidentified or at unknown locations in the visual system were observed in 30.7% and 9.3% of patients, respectively (Table 1) . The incidence of oscillatory response in unfiltered recordings was not correlated to variables such as visual acuity, field defect, age, and duration of disease, nor did it depend on the stimulus spatial frequency. No differences in the oscillatory response waveform or in the signal amplitude distribution were observed between the patients with prechiasmatic or postchiasmatic impairment or among subgroups of patients with different disorders or groups of disorders. Larger subgroups are necessary to further investigate the correlation with different disorders. 

Discussion

Tentatively understood as the result of synchronous, stimulus-related activation in the gamma range of neuronal assemblies responding to the visual properties of stimulus, ^{31 32 36 41 42 43 44 45 47 48} the 20–45 Hz human oscillatory mass response shares characteristics of conventional VEPs, such as the stimulus/response function for contrast and spatial frequency tuning, but does have differences, such as in time dynamics and source orientation in striate cortex. ^{14 23 28 33 36 57} The observation that oscillatory gamma activity originating in visual cortex ^{23 31 32} can be evoked by contrast stimulation in the absence of conventional VEPs further indicates some independence among these electrophysiological events and supports an early functional role of gamma band activity in the stimulus-induced neuronal synchronization and in visual information processing. It also implicates some functional hierarchy among brain signals, consistent with evidence from studies on the hippocampus and the auditory and visual cortices that oscillatory signals at different frequencies can trigger or modulate each other and with the oscillatory structure that appears to be intrinsic to brain signals and reflects the excitability of neuronal assemblies. ^{50 59 60 61 62 63 64 65 66 67} The extent to which VEP and oscillatory responses may mirror distinct parallel or consecutive phases/mechanisms of visual information processing remains speculative, yet a differential sensitivity to disorders of the visual system appears plausible. However, the processes generating the low-frequency VEP and gamma oscillations are too complex to allow detailed hypotheses about selective pathophysiological interferences on these responses. In this study, isolated or preponderant oscillatory responses were observed in unfiltered recordings in patients with either prechiasmatic or postchiasmatic damage, though with different incidences (60%–30.7%), and their occurrence does not appear related to any of the pathophysiological conditions considered or attributable to any single diagnosis or group of diagnoses. Instead, impaired transfer of visual information to cortex in readable formats or improper signal decoding appear functionally equivalent pathophysiological conditions irrespective of disease. 

Oscillatory responses that are phase-locked to the stimulus, comparable in morphology and frequency organization to those separated by high-pass filtering in healthy subjects, and unaffected by pathologic processes interfering with the mechanisms generating the VEP can result in abnormalities of the VEP waveform such as those previously described in patients with neurologic diseases. ^{7 9 11 12 13} Oscillatory responses can thus mimic a degraded VEP waveform, which has clinical relevance. The observation of a predominant oscillatory response in unfiltered VEP recordings would indicate inadequate organization of the cortical response and therefore indicate functional impairment irrespective of the latencies of recognizable waves. In this regard, it should be noted that several waves of the oscillatory response are superimposed in time to wave P100 of the VEPs in both healthy subjects ¹⁴ and patients and—if prominent on VEPs with reduced amplitude—may erroneously be labeled as wave P100 with normal or increased latency. Incorrect estimates may result from this misinterpretation and would reduce the reliability of electrophysiological visual testing, whereas appropriate identification of the (partially independent) oscillatory responses would improve diagnostic accuracy. 

 Supported by the University of Genova, Genova, Italy.

 Submitted for publication February 23, 2007; revised May 10, 2007; accepted August 27, 2007.

 Disclosure: W.G. Sannita, None; S. Carozzo, None; M. Fioretto, None; S. Garbarino, None; C. Martinoli, None

 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.

 

Superimposed unfiltered VEPs (top) and high-pass filtered (DFT filter; 19-Hz cutoff; bottom) oscillatory responses to contrast stimulation recorded from a healthy subject. The individual waves of the oscillatory response are labeled by polarity and order of appearance. ¹⁴ Right eye, monocular stimulation at 5 cyc/deg. The signal amplitude distribution before and after high-pass filtering is shown in the amplitude spectra on right column (graphically truncated at 150 Hz). Each averaged VEP or oscillatory response here and in Figures 2 3 4 5is based on 100 epochs free of artifacts.

Figure 2.

 

Patient (63 years old) with multiple high-intensity, irregularly enhancing MRI lesions involving the right temporal cortex and optic radiation, conceivably caused by strokes. (A) Conventional VEPs with increased P100 latency (127 ms) recorded after contrast stimulation of the right eye at 5 cyc/deg. (B) VEPs to stimulation of the left eye at the same spatial frequency are dominated by a sequence of waves (latency of first wave, 98 ms) in the absence of unambiguous wave P100. Note the two peaks in the amplitude spectrum, with the peak at 17.0 to 35 Hz prevailing on slow (0.5–17 Hz) frequencies (right, middle). (C) Oscillatory responses after high-pass filtering (B) at 19 Hz, with individual waves superimposed in time to those observed in unfiltered recordings. Two consecutive, unfiltered responses are shown (D).

Figure 3.

 

Patient (78 years old) with multiple deep infarction and cortical atrophy involving visual areas. Right eye stimulation at 1.3 cyc/deg. (A) VEPs with bifid (or W-shaped) P100 recorded from left occipital (latencies of the two major positive waves at 92 ms and 111 ms, respectively), and VEPs from right occipital dominated by a sequence of waves without unambiguous P100. (B) Symmetrical oscillatory responses were obtained after high-pass signal filtering at 19 Hz, with individual waves at latencies comparable to those observed in broadband recordings (A). (C) After removing the 19- to 35-Hz frequency components, VEPs with regular morphology (P100 at 97-ms latency) were recorded on left occipital, whereas no VEP was observed on right occipital.

Figure 4.

 

Signal amplitude spectra from four healthy controls (A) and six patients (B) whose unfiltered VEP recordings were dominated by the oscillatory responses. In all cases, right eye stimulation was at 5.0 cyc/deg and 80% contrast.

Figure 5.

 

Amplitude spectrum in the 0.5- to 250-Hz frequency interval for the group of patients with preponderant oscillatory response in unfiltered signals (gray) and the healthy controls (black). Mean (top) ± SE (bottom) across subjects. Same data after logarithmic transform (inset). Frequency resolution (1.03 Hz) and smoothing. Note the smaller amplitude in the 0.5- to 19-Hz frequency interval representative of VEPs in the patient groups (P < 0.0001 at each frequency component) in the absence of significant differences between groups in the approximately 19- to 35-Hz frequency interval.

Figure 6.

 

Time/frequency distribution of the signal response phase-locked or not phase-locked to stimulus in controls (left) and patients with predominant oscillatory response in unfiltered recordings (right). Top: time/frequency distribution of the signal activity independent of phase-locking (TSE). Middle: signal activity phase-locked to stimulus (RIA). Bottom: normalized LOI, that is, the signal activity phase-locked to stimulus (RIA) normalized versus the total signal activity independent of phase-locking (TSE). In controls, the phase-locked activity (RIA and LOI) clusters in two frequency intervals, at approximately 20 to 40 Hz (oscillatory response) and in the low-frequency components below 19 Hz (VEP), respectively; the observation replicates previous findings ^{28 57} and confirms the earlier time dynamics of the oscillatory response. In patients, clusters of activity phase-locked to stimulus are evident only in the frequency range of the oscillatory response (approximately 20–40 Hz) but not in the low-frequency interval of the VEP; no activity above noise is evident in the frequency range lower than 19 Hz, expressive of the conventional VEP response. TSE, RIA, and LOI activities at each frequency and time point are indicated in pseudocolors according to the scale on the right (top). Time dynamics (LOI) in the frequency intervals 20 to 40 Hz (oscillatory response; red) and below 19 Hz (VEP low-frequency components; black) after amplitude normalization are also shown (bottom) to exemplify the oscillatory response anticipating the VEPs in healthy controls; the 20 to 40 Hz component is comparable in controls and patients, whereas low-frequency components are at noise level in patients. Averages in controls and patients. See Sannita et al. ²⁸ and Sannita et al. ⁵⁷ for detailed information on methods.

Table 1.

 

Characterization by Diagnosis of Patients with Preponderant Oscillatory Responses

Primary Diagnosis	No. Patients	Percentage
Postchiasmatic disease	45	59.9
Migraine (with or without aura)	6	8.0
Multiple sclerosis with prevalent postchiasmatic lesions	13	17.3
Space-occupying lesions involving postchiasmatic structures	5	6.7
Head injury with prevalent postchiasmatic lesions	10	13.3
Partial epilepsy with simple seizures originating in visual cortex	1	1.3
Chronic cerebrovascular diseases involving visual cortex	10	13.3
Prechiasmatic disease	23	30.8
Optic neuropathy (monocular)	8	10.7
Retinitis pigmentosa	8	10.7
Thrombosis of retinal artery	3	4.0
Maculopathy	4	5.4
Progressive or permanent visual impairment of undefined etiology	7	9.3

The authors thank Livio Narici (University Tor Vergata, Roma, Italy) for providing an early version of the software used to compute TSE, RIA, and LOI measures.