- Email: [email protected]
Potentials evoked by chirp-modulated tones: a new technique to evaluate oscillatory activity in the auditory pathway
Clinical Neurophysiology 115 (2004) 699–709 www.elsevier.com/locate/clinph
Potentials evoked by chirp-modulated tones: a new technique to evaluate oscillatory activity in the auditory pathway J. Artieda*, M. Valencia, M. Alegre, O. Olaziregi, E. Urrestarazu, J. Iriarte Neurophysiology Section, Department of Neurology, Clı´nica Universitaria and Fundacio´n para Investigacio´n Me´dica Aplicada, Universidad de Navarra, Avenida Pı´o XII 36, 31008 Pamplona (Navarra), Spain Accepted 20 October 2003
Abstract Objective: Steady-state potentials are oscillatory responses generated by a rhythmic stimulation of a sensory pathway. The frequency of the response, which follows the frequency of stimulation, is maximal at a stimulus rate of 40 Hz for auditory stimuli. The exact cause of these maximal responses is not known, although some authors have suggested that they might be related to the ‘working frequency’ of the auditory cortex. Testing of the responses to different frequencies of stimulation may be lengthy if a single frequency is studied at a time. Our aim was to develop a fast technique to explore the oscillatory response to auditory stimuli, using a tone modulated in amplitude by a sinusoid whose frequency increases linearly in frequency (‘chirp’) from 1 to 120 Hz. Methods: Time-frequency transforms were used for the analysis of the evoked responses in 10 subjects. Also, we analyzed whether the peaks in these responses were due to increases of amplitude or to phase-locking phenomena, using single-sweep time-frequency transforms and inter-trial phase analysis. Results: The pattern observed in the time-frequency transform of the chirp-evoked potential was very similar in all subjects: a diagonal band of energy was observed, corresponding to the frequency of modulation at each time instant. Two components were present in the band, one around 45 Hz (30 –60 Hz) and a smaller one between 80 and 120 Hz. Inter-trial phase analysis showed that these components were mainly due to phase locking phenomena. Conclusions: A simultaneous testing of the amplitude-modulation-following oscillatory responses to auditory stimulation is feasible using a tone modulated in amplitude at increasing frequencies. The maximal energies found at stimulation frequencies around 40 Hz are probably due to increased phase-locking of the individual responses. q 2004 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. Keywords: Steady state evoked potential; Oscillatory activity; Auditory evoked potential; Human
1. Introduction Cortical oscillatory activity may be a key mechanism in perceptual binding (Singer, 1993). The role of synchronous oscillations in perception has been suggested by studies performed at very different levels, from single neuron activity to EEG and MEG (Engel et al., 1992; TallonBaudry et al., 1996; Rodriguez et al., 1999). These two latter techniques are limited by the spatial filter imposed by the head, although they have the invaluable advantage of being non-invasive. There are several approaches to the study of cortical oscillatory activity using EEG. An oscillatory response to * Corresponding author. Tel.: þ 34-948-255400; fax: þ 34-948-296500. E-mail address: [email protected] (J. Artieda).
a stimulus may have a constant phase relationship with it (phase-locked or ‘evoked’ response) or may consist of a change in the amplitude in the ongoing activity without a clear phase relationship with the stimulus (non phase-locked or ‘induced’ response) (Pfurtscheller and Lopes da Silva, 1999). The most typical examples of phase-locked responses are the evoked potentials. There are two possible mechanisms that can generate an evoked potential in an averaged response. It can result from the sum of lowamplitude potentials generated in each individual sweep; with the averaging process, the amplitude of the background EEG noise is reduced and these low-amplitude potentials come up and can be clearly observed. Alternatively, these responses might be due to synchronous phase-resetting in the ongoing activity caused by the stimulus, without any
1388-2457/$30.00 q 2004 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.clinph.2003.10.021
700
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
amplitude changes (Bas¸ar et al., 1992; Makeig et al., 2002; Penny et al., 2002). Steady-state responses (SSRs) are the result of averaging individual responses to trains of rhythmic stimuli delivered at a constant frequency (so the ‘steady-state’ name) (Stapells et al., 1984). They have a frequency content which is maximum at the frequency of stimulation. SSRs have been obtained with auditory and visual stimuli. Auditory steady-state responses were initially described with trains of clicks (Galambos et al., 1981), but they can also be obtained with amplitude-modulated (AM) tones (Picton et al., 1987). The amplitude of the oscillations is maximal at stimulation frequencies around 40 Hz (30 –50 Hz) (Galambos et al., 1981), with a second component of smaller amplitude between 80 and 120 Hz (Lins et al., 1995). The cause of the maximal response at stimulation rates around 40 Hz has been intensely debated; it is not clear whether it is just due to the superposition of the middle latency responses to each single stimulus or there is some kind of resonance phenomena in the auditory pathway at that frequency (John and Picton, 2000). In this latter case, the frequency of the maximal response could indicate the preferential working frequency of the auditory network. An alteration in the brain ‘working frequencies’ has been hypothesized as a possible cause of some neurological disorders (McCormick, 1999). If one modulation frequency is used at a time, testing the SSRs to different stimulation frequencies can be very long, depending on the number of frequencies to examine. A faster approach has been proposed, both for visual and auditory steady-state responses, with the help of a Fourier analyzer (Regan, 1977; Stapells et al., 1984). Our purpose was to prove the feasibility of a test that allows the simultaneous study of the auditory steady-state responses to many different frequencies of stimulation, using a tone modulated in amplitude at increasing frequencies from 1 to 120 Hz (‘chirp’). As a secondary endpoint, our study tried to establish whether the higher amplitude of the frequency-following responses around 40 Hz is due to an increase in the amplitude of the response at this frequency or to an increase in the inter-trial phaselocking.
2. Subjects and methods Twelve young healthy subjects were initially included in the study. All of them were informed in detail about the experiment, and gave their written consent. Two subjects were excluded ‘a posteriori’ from the analysis, one of them due to a previously undocumented partial hearing loss in the range of the carrier tone, and another due to a highamplitude muscle artifact. Therefore, only 10 subjects were included in the final analysis. The protocol was approved by the institutional ethics committee.
Thirteen EEG channels were recorded in monopolar montage referred to both earlobes (F3, Fz, F4, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, according to the 10– 20 International system (Jasper, 1958)), using a commercial electrode cap (Electro-CAP) and Bio-Logic amplifiers. The signal was amplified £20 000; filtered 1– 300 Hz, digitized at 750 Hz and stored in a PC for offline analysis. A minimum of 500 sweeps were recorded per subject, synchronized with the beginning of the sound. The length of the sweep was 2.09 s, with a pre-stimulus period of 0.16 s. The auditory stimuli were designed in the Matlab environment (Mathworks, USA), and delivered through the STIM module of the Neuroscan system (NeuroSoft, El Paso, USA). Bilateral intracanalicular earphones were used. The stimulus consisted of a 85 dB HL 1200 Hz tone modulated in amplitude by a linear chirp, a sinusoid whose frequency increased linearly from 1 to 120 Hz (see Fig. 1). The total length of the sound was 1.61 s. Some additional tests were carried out in a small subgroup of subjects to discard distortions in the response caused by the parameters of the modulation rate. In two of them, longer and shorter sounds (with faster and slower increase of the modulation rate, respectively) were tested and compared with the ‘standard’ responses. Also, in another two subjects, a decreasing-rate chirp modulation (of the same length and frequency range) was used. Finally, in two additional subjects the responses were recorded both in awake and sleep states, using the standard stimulation sound. Three different mathematical approaches were applied to the recorded signal in each experiment. 2.1. Time-frequency transform of the average All sweeps from each subject and condition were averaged, in order to obtain the chirp-evoked potentials. Steady-state potentials are usually studied by means of Fourier transforms. The fast Fourier transform (FFT) gives an accurate estimation of the power spectrum of a stationary signal. However, the responses evoked by a variably modulated tone should not be stationary by definition. Time-frequency transforms can give an estimation of the time-varying energy of a signal in each frequency band (Cohen, 1989). A time-frequency analysis (Gabor transform) was applied to the averaged signal (chirp-evoked potential) from each subject (Gurtubay et al., 2001). The results from all individual subjects were also grand-averaged and displayed using a 3-dimensional (3D) color graph. The same procedure was applied to the modulation of the tone (chirp), in order to assure a constant energy content through time and frequency. 2.2. Average of single-sweep time-frequency transforms Time-frequency transforms are typically used to represent energy values, which are always positive despite
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
701
Fig. 1. Illustration of the procedure followed to build the stimulus. A 1200 Hz tone was modulated in amplitude with a linear chirp, a sinusoid whose frequency increased linearly from 1 to 120 Hz.
the phase. Therefore, the average of single-sweep timefrequency transforms yields an estimation of the energy per sweep taking only into account amplitude changes, and not inter-sweep phase interactions (Tallon-Baudry et al., 1996; Gurtubay et al., 2001). An increase in this graph indicates then that there is an amplitude increase in a frequency at a time point, which is constant in most sweeps. As in the previous analysis, the results were displayed for each individual subject in a 3D colored graph. Also, the average of the transforms was normalized (divided) by the mean energy for each frequency during the pre-stimulus period.
2.3. Inter-trial coherence Any oscillatory activity at a frequency can be defined by two components: amplitude and phase. The procedure outlined above (average of single-sweep time-frequency transforms) yields an estimation of the inter-trial constancy of amplitude changes. As opposed to this analysis, the intertrial coherence (Makeig et al., 2002), also known as phaselocking factor (Tallon-Baudry et al., 1996), gives an estimation of the inter-trial constancy of the phase distribution (see Appendix A). According to this analysis, we obtain a parameter that ranges from 0 (for purely non phase-locked activity) to 1 (for strictly phase-locked activity). The results were displayed in a 3D graph, as in the previous analysis.
3. Results 3.1. Steady-state responses The average of the responses yielded in all subjects an oscillatory potential with several components (Fig. 2, top left). Although some oscillatory changes were already present in some subjects in partial averages of as few as 30 sweeps, the signal-to noise ratio and the amplitude of the changes increased steadily up to 400– 500 sweeps per subject. An initial response, consisting of two consecutive N1-P2 complexes, was observed about 120 ms after the beginning of the sound (mean latency for the first N1 peak: 120, range: 82 –143 ms; mean latency for the second N1 peak: 280 ms). A small biphasic wave was present before it, corresponding to the transient gamma band response (see below). 300 ms after the sound, oscillations at increasing frequencies were observed. The highest amplitude of these oscillations was recorded about 600 ms after the beginning of the sound, what corresponded to a modulation frequency of 45 Hz in the stimulus. The maximum amplitude of the oscillatory response ranged between 2 and 7 mV. A decrease in amplitude was observed in most cases between 900 and 1100 ms, with a late increase between this time and the end of the stimulus (1600 ms). At the end of the stimulus, an offset potential was observed (mean latencies of 56 ms for the negative (N10 ) and 145 ms for the positive peak (P20 ), calculated after the end of the stimulus). A slow positive wave (1 Hz) overlapped the initial part of the oscillatory component.
702
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
Fig. 2. (Left) Grand average of the potential evoked by the chirp-modulated tone (top), and grand average of the time-frequency transforms of the potentials from each subject (bottom), in Fz. The color scale represents absolute energy values. Time 0 corresponds to the beginning of the stimulus. (Top right) Topographical representation of the energy contained in two time-frequency areas, representative of the two main components in the response. (Bottom right) Time-frequency energy distribution of the modulation signal (‘chirp’). Energy values are constant through the diagonal band that increases linearly in frequency along time. The color scale represents relative values.
3.2. Oscillatory activity: time-frequency transform of the evoked potential The results were highly consistent in the whole group of 10 subjects, both in absolute energy values and in energy distribution. The main finding in the plots was a diagonal band of energy, that represented the oscillatory response at the increasing frequency of modulation (Fig. 2, bottom left). This response began approximately at a stimulation frequency of 20 Hz and lasted till the end of the stimulus (120 Hz). The maximum energy was at frequencies around 45 Hz (mean: 45.9, range: 43 –50). In most subjects, two components could be observed in this response, one between 20 and 65 Hz, with higher energy, and a second component between 80 and 120 Hz. The energy content in intermediate frequencies (65 – 80 Hz) was usually much lower. The topography of the lower component (20 –65 Hz) was markedly frontal mesial in the initial montage referred to both earlobes; after computing laplacian derivatives, the maximum energy shifted to both temporal regions. In the parietal and posterior temporal electrodes, the maximum relative amplitude of the response was at 60 Hz, although with lower energy than in frontal leads. The topography of the second component was more diffuse, although it also had frontal preponderance. In half of the subjects, and in the grand average, a second diagonal was observed (with much lower
energy values), which corresponded to a response at twice the stimulation frequency. A small energy peak around 45 Hz was also observed shortly (30 – 35 ms) after the beginning of the stimulus. The frequency of this peak was similar to that of the maximum energy. This peak corresponded to the transient gamma activity evoked by the initial tone (Pantev et al., 1991). A linear representation of the energy changes in representative frequencies of the first and second components of the oscillatory response is shown in Fig. 3; the 40 Hz transient response can be observed in the 45 Hz plot (arrow). Fig. 4 shows the global energy content of the stimulus and the averaged response; the two oscillatory components can be clearly observed in the latter. 3.3. Average of single-sweep time-frequency transforms: amplitude changes The main finding in all subjects was a progressive decrease in energy for higher frequencies during the whole period analyzed, due at least in part to the low-pass filter effect of the skull. Stimulus-related changes were minimal. A ‘diagonal’ band corresponding to the modulation frequency was observed only in two of the subjects (one of them shown in Fig. 5; maximum increase of 60% over baseline). In another 3 subjects, a small increase in energy was observed around 40 Hz only after the normalization of
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
703
Fig. 3. Linear representation of energy changes in Fz along time at two different frequencies, representative of the two main components in the oscillatory response. The dotted line represents energy changes in the modulation signal (chirp), in relative values (right scale). The vertical arrow indicates the 40 Hz transient response. The main peak in both plots matches the frequency of stimulation; the smaller peak observed earlier is a response at twice the frequency of stimulation. Time 0 corresponds to the beginning of the stimulus.
the graph with the pre-stimulus period, at the same time point observed in the time-frequency plot of the averaged potential. Also, a 50 Hz band (mains frequency) was observed in some plots, despite the effort to keep low impedances. Although interindividual variation in absolute energy values was high (as in any non-averaged EEG recording), the proportion of energy in different frequencies was similar.
3.5. Effects of the chirp length and direction The results obtained with larger sweeps (with a 2.5 times slower increase rate of the modulation frequency) were comparable to the results obtained with our ‘standard’ chirp. The increase in the signal-to-noise ratio was minimal,
3.4. Inter-trial coherence/phase-locking factor The results were in all subjects very similar in general shape to those observed in the time-frequency transform of the average. A marked diagonal band was observed in all cases, representing the response to the modulation frequency. Two peaks were observed around 45 and 90 Hz (the same frequencies as in the previous graph). The distribution of the phase-locking factor throughout frequencies matched the energy changes in the TF transform of the average. Also, a small 40 Hz peak corresponding to the 40 Hz transient could be observed in many subjects. The phase-locking values (. 0.4 in all subjects) were similar for the two components of the response (30 – 65 Hz and 80 – 120 Hz). These values were highly significant according to the statistical procedure described by Feige et al. (2000). Fig. 5 includes an example of the 3 mathematical procedures applied to the same data from one representative subject.
Fig. 4. (A) Estimate of the energy spectrum of the modulation signal in the range studied, obtained by adding all energy values at each frequency through the period analyzed. Values are given in relative units. (B) Estimate of the energy spectrum of the grand-averaged response in the range studied, obtained by adding all energy values at each frequency through the period analyzed. Two peaks, corresponding to the main components of the oscillatory response (45 and 95 Hz), can be observed.
704
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
Fig. 5. Time-frequency plots of the 3 mathematical procedures applied to the data from a single representative subject. (Top left) Energy distribution of the averaged response. The color scale represents absolute energy values. (Top right) Phase-locking factor from the same subject. Values can range from 0 to 1 (see text). (Bottom left) Average of single-sweep time-frequency transforms from the same subject. The color scale represents absolute energy values. (Bottom right) Normalization with baseline energy of the bottom left graph. The color scale indicates percentual relative increases or decreases at each frequency compared with the pre-stimulus period. A small band of energy increase can be observed more clearly.
Fig. 6. Responses to an increasing-rate (left) and a decreasing-rate (right) modulated tone in a single subject. The averaged potential (top) and its timefrequency energy distribution (bottom) are plotted at the same scale in both conditions. The energy distribution in frequency in both TF plots is nearly identical, although inverted in time.
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
705
Fig. 7. Responses to a chirp-modulated tone in a single subject during wakefulness (left) and stage 2 sleep (right). The averaged potential (top) and its timefrequency energy distribution (bottom) are plotted at the same scale in both conditions. A description of the changes found during sleep can be found in the main text.
Fig. 8. Averaged responses (top), time-frequency energy distribution (middle), and phase-locking factor (PLF, bottom) obtained in one subject after averaging different numbers of sweeps. Please note that although PLF values apparently decrease when more sweeps are included, the statistical significance increases steadily with an inverse exponential relationship with the number of sweeps. Thus, P values in the low gamma range are hardly significant in the first plots (P , 0:05), and highly significant in the last ones (P , 0:00001). The recording time for each plot is indicated below them.
706
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
despite a much longer duration of the test. Some differences in relative energy values were present in the lower part of the frequency range studied (around 20 –25 Hz) due to the slower increase of the modulation rate, but the absolute maximal values and the distribution of the energy changes in the 30 –120 Hz range were similar. With shorter sweeps (two times faster increase rate of the modulation frequency), the results were much worse in terms of signal-to noise ratio. In some subjects, instead of a diagonal band only isolated peaks of response could be identified in the TF energy distribution. There were no significant differences in the frequency distribution of the energy peaks between the oscillatory responses obtained with an increasing-rate chirp and a decreasing-rate chirp modulation (see Fig. 6), with the exception of a larger 40 Hz transient with the decreasingchirp stimulation. In particular, the frequency of maximal response in both components (around 45 and 90 Hz) in each subject were the same. 3.6. Effects of sleep In the two subjects studied, the amplitude of the responses was smaller during sleep (stage 2), for both components of the response. The decrease in energy was higher for the 40 Hz component of the response (1/6 of awake values) than for the 80 Hz component (1/4). The maximum amplitude of the response was at a slightly lower frequency during the sleep state (36 vs. 41 Hz). Fig. 7 compares the results obtained during sleep and wakefulness in one subject.
4. Discussion 4.1. Chirp-evoked potentials as a new test Our results have shown that it is possible to analyze the oscillatory response to different frequencies of stimulation in a simultaneous way using the time-frequency transform of the potential evoked by a chirp-modulated tone (a tone modulated in amplitude at increasing frequencies). With this technique, the frequency at which the maximum response is obtained, as well as the range of frequencies in which there is an oscillatory response, can be easily determined. The results obtained are comparable to those described in the literature using much longer procedures. A maximum of 500 sweeps of 2.09 s of length (17 min) were required for a good definition of the energy changes in all subjects (see Fig. 8). At least 4 components could be observed in the response. First, an ‘onset’ complex was observed, consisting of the transient 40 Hz response and two long-latency auditory evoked potentials evoked by the two first ‘beats’ (modulation cycles) of the sound. The longer latency of the first long-latency auditory evoked potential and
the transient 40 Hz response compared with other studies can be explained by the modulation of the tone, that caused a very slow rising slope in the first cycle (thus delaying the reaching of the threshold intensity). An ‘offset’ response (Noda et al., 1998) could be also observed in the averaged potential, without a clear correlation with frequency changes in the range studied. Finally, two clear components were observed in the oscillatory response to the frequency of modulation. The first component, with the highest energy, covered the range between 30 and 65 Hz in most subjects. The second component was between 80 and 120 Hz. The origin of these two components is probably very different, according to the literature. The low gamma response (30 – 65 Hz) had a maximum peak around 45 Hz. The frequency of this peak did not differ from the frequency of the transient gamma activity found in a previous study by our group (Gurtubay et al., 2001). This response corresponds to the initial description of the steady-state responses (Galambos et al., 1981). Several studies have located the origin of this response in the auditory cortex (Romani et al., 1982; Makela and Hari, 1987; Makela et al., 1990), but at a location different to the origin of the 40 Hz transient (Pantev et al., 1993). Other studies suggest, however, that there may also be some contribution of brain-stem generators to this response (Hori et al., 1993; Tsuzuku, 1993). The high gamma responses (80 – 120 Hz), on the contrary, are probably mainly originated in the brain-stem (Herdman et al., 2002). We found a maximal response in the low gamma range over Fz and F3, as described in other studies using the same earlobe references (Pastor et al., 2002). The maximum response in the high gamma range had also a frontal predominance. Even though the number of channels used in the present study can not yield a detailed topography, our data are concordant with the literature, suggesting that the responses obtained are equivalent to those described with equivalent constantly-modulated tones. The fact that oscillatory responses at the frequency of stimulation can be obtained with stimuli delivered at variable rates makes the term ‘amplitude modulationfollowing responses (AMFR)’ (Levi et al., 1993) preferable to the term ‘steady-state responses,’ more widely used. The results obtained in the sleep state are fully concordant with previously published data on amplitude changes, thus supporting the validity of our technique to assess the size of the responses at different stimulation frequencies. 4.2. Amplitude and phase Our results have confirmed the known fact that our brain responds to an amplitude-modulated tone with oscillations at the frequency of modulation, within a frequency range. This response is clearly higher for frequencies slightly higher than 40 Hz. A higher amplitude in an average can be due either to higher amplitude in the individual responses, to higher phase constancy in the individual responses, or to
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
a combination of both. In the initial description of the 40 Hz steady-state response, Galambos et al. (Galambos et al., 1981) suggested that this response might be due to a superposition of the middle latency responses evoked by each of the clicks in the series. The results obtained in the phase-locking index clearly indicate that there is a higher constancy in the phase at 40 Hz than at other frequencies of stimulation, thus indicating that phase adjustment is responsible for most of this phenomenon. However, a higher amplitude in the individual response at 40 Hz was also suggested by the small peak in the average of the individual time-frequency transforms, so a higher amplitude in the individual responses to a stimulation at 40 Hz may also be involved. Although the higher signal-to-noise ratio generated by an increment in amplitude might cause an apparent increased phase-synchronization in the phase-locking index plot, the fact that this amplitude increment was only observed in some subjects, and its small size suggests that the findings of phase-synchronization are real. Besides, a higher amplitude in an EEG wave may be caused by higher phase synchronization between the neuronal groups located in the nearby regions. The differences in energy observed in the Gabor transform of the averaged potentials between the low gamma and the high gamma component of the oscillatory response were small or even absent both in the normalized average of the individual Gabor transforms and in the inter-trial coherence plot. The inter-trial coherence plot only represents phase phenomena, without influence of amplitude changes. The average of the Gabor transforms only includes amplitude information, but the normalization corrects for the baseline energy for each frequency, thus giving relative values. This might indicate that the higher energy observed in the low gamma component might be at least partially due to the low-pass filter effect of the skull (Pfurtscheller and Cooper, 1975).
707
synaptic activity at this frequency of stimulation (Pastor et al., 2002). Also, a marked cerebellar activation was also observed only at this frequency. The authors suggested that the cerebellum might be ‘braking’ the extension of the ‘resonant’ cortical activity at this frequency. Our results also suggest that the maximal 40 Hz response may be related to an increased synchronization through the individual responses at this stimulation rate. Why the auditory pathway tends to synchronize better at 40 Hz than at other frequencies? One possible answer is because 40 Hz is one of its usual ‘working’ frequencies. Gamma oscillations may be involved in perceptual binding, as shown in studies with visual or auditory stimulation (Tallon-Baudry et al., 1996; Gurtubay et al., 2001). Some authors have proposed that alterations in the thalamo-cortical circuits involved in the generation of rhythmic gamma activities might be involved in the origin of some neurological diseases (Llinas et al., 1999; McCormick, 1999). There are several descriptions of altered responses to steady-state auditory or visual stimulation in neurological or psychiatric diseases (Butler et al., 2001). A displacement towards lower frequencies of the maximum response to auditory steady-state stimuli has been described in Parkinson’s disease (Pastor and Artieda, 1996). The variability observed in the maximum energy peak was minimal in our study. Therefore, our method provides a fast and reliable test to determine the range of frequencies and the frequency of maximal response to auditory rhythmic stimulation that can be used for the comparison between normal subjects and different pathologies.
Acknowledgements This work has been supported by grant 030043 from the Fondo de Investigaciones Sanitarias, Spain.
4.3. The meaning of the 40 Hz response There are several theories on the origin of the 40 Hz response. As commented above, it was initially thought that it was due to a superposition of the auditory middle-latency evoked potentials (Galambos et al., 1981; Hari et al., 1989). Some recent studies using a multiple generator model support this idea (Gutschalk et al., 1999). However, the persistence of the steady-state responses after the end of the stimulus indicate that resonance phenomena (Bas¸ar et al., 1987) might contribute to this response (Santarelli et al., 1995). Our results do not support either the superposition theory, as we obtained the same results (maximal responses around 40 Hz) using a variable modulation, that implies a progressively shorter cycle hampering the superposition of waves. A recent study using PET showed that there is a higher activation of the auditory cortex during a 40 Hz steady-state stimulation than with other frequencies, indicating a higher
Appendix A The degree of phase resetting (i.e. phase consistency or phase locking) of EEG activity in single trials can be estimated with the procedure proposed by Tallon-Baudry et al. (1996). The method yields a parameter named phaselocking factor, also known as inter-trial coherence (Makeig, 2002), that gives an estimation of the inter-trial constancy of the phase distributions. For each trial, we compute its convolution with a complex wavelet centered in the frequencies of interest. Here we have used a Morlet wavelet
wtðtf Þ ¼ e
2
t2 2s2t
ej2pft
708
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
with
st ¼
1 2psf
The wavelet family is characterized the constant ratio: Nh0 ¼ st ·f ; we have used Nh0 ¼ 10: Next, we extract the phase of this convolution fðt; f ; nÞ for every time instant t, frequency f and trial n ¼ ½1; …; N: The inter-trial coherence value is then defined for every time and frequency bin as the average across trials. ccomplex ðt; f Þ ¼
N 1 X ejfðt;f ;nÞ N n¼1
The amplitude c of the resulting average represents a measure of the uniformity of the distribution of phases, i.e. how consistent the phase resettings across trials are. In order to provide a statistical significance level to the inter-trial coherence values, we have used the approximation proposed by Feige et al. (2000). Their work was focused on phase differences; our purpose is similar, as we try to measure how far from a uniform distribution the calculated phase distributions are. According to this test, the statistical significance of a c inter-trial coherence value determined from N trials can be calculated as P ¼ e2Nc
2
where P, for a given value of ITC c calculated from N trials, describes the probability to obtain an equal or larger ITC value by chance.
References Bas¸ar E, Rosen B, Bas¸ar-Eroglu C, Greitschus F. The associations between 40 Hz-EEG and the middle latency response of the auditory evoked potential. Int J Neurosci 1987;33:103–17. Bas¸ar E, Bas¸ar-Eroglu C, Parnefjord R, Rahn E, Schurmann M. Evoked potentials: ensembles of brain induced rhythmicities in the alpha, theta and gamma ranges. In: Bas¸ar E, Bullock TH, editors. Induced rhythms in the brain. Brain dynamic series, Boston, MA: Birkhauser; 1992. p. 155 –81. Butler PD, Schechter I, Zemon V, Schwartz SG, Greenstein VC, Gordon J, Schroeder CE, Javitt DC. Dysfunction of early-stage visual processing in schizophrenia. Am J Psychiatry 2001;158:1126–33. Cohen L. Time-frequency distributions. A review. Proc IEEE 1989;77: 941– 81. Engel AK, Konig P, Kreiter AK, Schillen TB, Singer W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci 1992;15:218 –26. Feige B, Aertsen A, Kristeva-Feige R. Dynamic synchronization between multiple cortical motor areas and muscle activity in phasic voluntary movements. J Neurophysiol 2000;84:2622 –9. Galambos R, Makeig S, Talmachoff PJ. A 40-Hz auditory potential recorded from the human scalp. Proc Natl Acad Sci USA 1981;78: 2643–7. Gurtubay IG, Alegre M, Labarga A, Malanda A, Iriarte J, Artieda J. Gamma band activity in an auditory oddball paradigm studied with the wavelet transform. Clin Neurophysiol 2001;112:1219–28.
Gutschalk A, Mase R, Roth R, Ille N, Rupp A, Hahnel S, Picton TW, Scherg M. Deconvolution of 40 Hz steady-state fields reveals two overlapping source activities of the human auditory cortex. Clin Neurophysiol 1999;110:856–68. Hari R, Hamalainen M, Joutsiniemi SL. Neuromagnetic steady-state responses to auditory stimuli. J Acoust Soc Am 1989;86:1033 –9. Herdman AT, Lins O, Van Roon P, Stapells DR, Scherg M, Picton TW. Intracerebral sources of human auditory steady-state responses. Brain Topogr 2002;15:69– 86. Hori A, Yasuhara A, Naito H, Yasuhara M. Steady-state auditory evoked potentials (SSAEPs) in the rabbit. Contribution of the inferior colliculus. Electroenceph clin Neurophysiol 1993;88:229–36. Jasper HH. Report of the Committee on Methods of Clinical Examination in Electroencephalography. Electroenceph clin Neurophysiol 1958;10: 370 –5. John MS, Picton TW. Human auditory steady-state responses to amplitudemodulated tones: phase and latency measurements. Hear Res 2000;141: 57 –79. Levi EC, Folsom RC, Dobie RA. Amplitude-modulation following response (AMFR): effects of modulation rate, carrier frequency, age, and state. Hear Res 1993;68:42– 52. Lins OG, Picton PE, Picton TW, Champagne SC, Durieux-Smith A. Auditory steady-state responses to tones amplitude-modulated at 80– 110 Hz. J Acoust Soc Am 1995;97:3051 –63. Llinas RR, Ribary U, Jeanmonod D, Kronberg E, Mitra PP. Thalamocortical dysrhythmia: a neurological and neuropsychiatric syndrome characterized by magnetoencephalography. Proc Natl Acad Sci USA 1999;96:15222–7. Makeig S, Westerfield M, Jung TP, Enghoff S, Townsend J, Courchesne E, Sejnowski TJ. Dynamic brain sources of visual evoked responses. Science 2002;295:690– 4. Makela JP, Hari R. Evidence for cortical origin of the 40 Hz auditory evoked response in man. Electroenceph clin Neurophysiol 1987;66:539–46. Makela JP, Karmos G, Molnar M, Csepe V, Winkler I. Steady-state responses from the cat auditory cortex. Hear Res 1990;45:41– 50. McCormick DA. Are thalamocortical rhythms the Rosetta Stone of a subset of neurological disorders? Nat Med 1999;5:1349–51. Noda K, Tonoike M, Doi K, Koizuka I, Yamaguchi M, Seo R, Matsumoto N, Noiri T, Takeda N, Kubo T. Auditory evoked off-response: its source distribution is different from that of on-response. Neuroreport 1998;9: 2621–5. Pantev C, Makeig S, Hoke M, Galambos R, Hampson S, Gallen C. Human auditory evoked gamma-band magnetic fields. Proc Natl Acad Sci USA 1991;88:8996–9000. Pantev C, Elbert T, Makeig S, Hampson S, Eulitz C, Hoke M. Relationship of transient and steady-state auditory evoked fields. Electroenceph clin Neurophysiol 1993;88:389–96. Pastor MA, Artieda J. Cortical oscillatory behaviour to auditory repetitive stimulation in Parkinson’s disease. Mov Disord 1996;11(Suppl):128. Pastor MA, Artieda J, Arbizu J, Marti-Climent JM, Penuelas I, Masdeu JC. Activation of human cerebral and cerebellar cortex by auditory stimulation at 40 Hz. J Neurosci 2002;22:10501–6. Penny W, Kiebel S, Kilner J, Rugg M. Event-related brain dynamics. Trends Neurosci 2002;25:287–9. Pfurtscheller G, Cooper R. Frequency dependence of the transmission of the EEG from cortex to scalp. Electroenceph clin Neurophysiol 1975; 38:93–6. Pfurtscheller G, Lopes da Silva FH. Functional meaning of event-related desynchronization (ERD) and synchronization (ERS). In: Pfurtscheller G, Lopes da Silva FH, editors. Event-related desynchronization. Handbook of electroencephalography and clinical neurophysiology. Revised series, Amsterdam: Elsevier; 1999. p. 51 –65. Picton TW, Skinner CR, Champagne SC, Kellett AJ, Maiste AC. Potentials evoked by the sinusoidal modulation of the amplitude or frequency of a tone. J Acoust Soc Am 1987;82:165–78. Regan D. Evoked potentials in basic and clinical research. In: Remond A, editor. EEG informatics. A didactic review and applications of EEG
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709 data processing. Amsterdam: Elsevier/North Holland Biomedical; 1977. p. 319–46. Rodriguez E, George N, Lachaux JP, Martinerie J, Renault B, Varela FJ. Perception’s shadow: long-distance synchronization of human brain activity. Nature 1999;397:430–3. Romani GL, Williamson SJ, Kaufman L. Tonotopic organization of the human auditory cortex. Science 1982;216:1339–40. Santarelli R, Maurizi M, Conti G, Ottaviani F, Paludetti G, Pettorossi VE. Generation of human auditory steady-state responses (SSRs). II. Addition of responses to individual stimuli. Hear Res 1995;83: 9–18.
709
Singer W. Synchronization of cortical activity and its putative role in information processing and learning. Annu Rev Physiol 1993;55: 349– 74. Stapells DR, Linden D, Suffield JB, Hamel G, Picton TW. Human auditory steady state potentials. Ear Hear 1984;5:105–13. Tallon-Baudry C, Bertrand O, Delpuech C, Pernier J. Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human. J Neurosci 1996;16:4240 –9. Tsuzuku T. 40-Hz steady state response in awake cats after bilateral chronic lesions in auditory cortices or inferior colliculi. Auris Nasus Larynx 1993;20:263–74.
Potentials evoked by chirp-modulated tones: a new technique to evaluate oscillatory activity in the auditory pathway J. Artieda*, M. Valencia, M. Alegre, O. Olaziregi, E. Urrestarazu, J. Iriarte Neurophysiology Section, Department of Neurology, Clı´nica Universitaria and Fundacio´n para Investigacio´n Me´dica Aplicada, Universidad de Navarra, Avenida Pı´o XII 36, 31008 Pamplona (Navarra), Spain Accepted 20 October 2003
Abstract Objective: Steady-state potentials are oscillatory responses generated by a rhythmic stimulation of a sensory pathway. The frequency of the response, which follows the frequency of stimulation, is maximal at a stimulus rate of 40 Hz for auditory stimuli. The exact cause of these maximal responses is not known, although some authors have suggested that they might be related to the ‘working frequency’ of the auditory cortex. Testing of the responses to different frequencies of stimulation may be lengthy if a single frequency is studied at a time. Our aim was to develop a fast technique to explore the oscillatory response to auditory stimuli, using a tone modulated in amplitude by a sinusoid whose frequency increases linearly in frequency (‘chirp’) from 1 to 120 Hz. Methods: Time-frequency transforms were used for the analysis of the evoked responses in 10 subjects. Also, we analyzed whether the peaks in these responses were due to increases of amplitude or to phase-locking phenomena, using single-sweep time-frequency transforms and inter-trial phase analysis. Results: The pattern observed in the time-frequency transform of the chirp-evoked potential was very similar in all subjects: a diagonal band of energy was observed, corresponding to the frequency of modulation at each time instant. Two components were present in the band, one around 45 Hz (30 –60 Hz) and a smaller one between 80 and 120 Hz. Inter-trial phase analysis showed that these components were mainly due to phase locking phenomena. Conclusions: A simultaneous testing of the amplitude-modulation-following oscillatory responses to auditory stimulation is feasible using a tone modulated in amplitude at increasing frequencies. The maximal energies found at stimulation frequencies around 40 Hz are probably due to increased phase-locking of the individual responses. q 2004 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. Keywords: Steady state evoked potential; Oscillatory activity; Auditory evoked potential; Human
1. Introduction Cortical oscillatory activity may be a key mechanism in perceptual binding (Singer, 1993). The role of synchronous oscillations in perception has been suggested by studies performed at very different levels, from single neuron activity to EEG and MEG (Engel et al., 1992; TallonBaudry et al., 1996; Rodriguez et al., 1999). These two latter techniques are limited by the spatial filter imposed by the head, although they have the invaluable advantage of being non-invasive. There are several approaches to the study of cortical oscillatory activity using EEG. An oscillatory response to * Corresponding author. Tel.: þ 34-948-255400; fax: þ 34-948-296500. E-mail address: [email protected] (J. Artieda).
a stimulus may have a constant phase relationship with it (phase-locked or ‘evoked’ response) or may consist of a change in the amplitude in the ongoing activity without a clear phase relationship with the stimulus (non phase-locked or ‘induced’ response) (Pfurtscheller and Lopes da Silva, 1999). The most typical examples of phase-locked responses are the evoked potentials. There are two possible mechanisms that can generate an evoked potential in an averaged response. It can result from the sum of lowamplitude potentials generated in each individual sweep; with the averaging process, the amplitude of the background EEG noise is reduced and these low-amplitude potentials come up and can be clearly observed. Alternatively, these responses might be due to synchronous phase-resetting in the ongoing activity caused by the stimulus, without any
1388-2457/$30.00 q 2004 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.clinph.2003.10.021
700
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
amplitude changes (Bas¸ar et al., 1992; Makeig et al., 2002; Penny et al., 2002). Steady-state responses (SSRs) are the result of averaging individual responses to trains of rhythmic stimuli delivered at a constant frequency (so the ‘steady-state’ name) (Stapells et al., 1984). They have a frequency content which is maximum at the frequency of stimulation. SSRs have been obtained with auditory and visual stimuli. Auditory steady-state responses were initially described with trains of clicks (Galambos et al., 1981), but they can also be obtained with amplitude-modulated (AM) tones (Picton et al., 1987). The amplitude of the oscillations is maximal at stimulation frequencies around 40 Hz (30 –50 Hz) (Galambos et al., 1981), with a second component of smaller amplitude between 80 and 120 Hz (Lins et al., 1995). The cause of the maximal response at stimulation rates around 40 Hz has been intensely debated; it is not clear whether it is just due to the superposition of the middle latency responses to each single stimulus or there is some kind of resonance phenomena in the auditory pathway at that frequency (John and Picton, 2000). In this latter case, the frequency of the maximal response could indicate the preferential working frequency of the auditory network. An alteration in the brain ‘working frequencies’ has been hypothesized as a possible cause of some neurological disorders (McCormick, 1999). If one modulation frequency is used at a time, testing the SSRs to different stimulation frequencies can be very long, depending on the number of frequencies to examine. A faster approach has been proposed, both for visual and auditory steady-state responses, with the help of a Fourier analyzer (Regan, 1977; Stapells et al., 1984). Our purpose was to prove the feasibility of a test that allows the simultaneous study of the auditory steady-state responses to many different frequencies of stimulation, using a tone modulated in amplitude at increasing frequencies from 1 to 120 Hz (‘chirp’). As a secondary endpoint, our study tried to establish whether the higher amplitude of the frequency-following responses around 40 Hz is due to an increase in the amplitude of the response at this frequency or to an increase in the inter-trial phaselocking.
2. Subjects and methods Twelve young healthy subjects were initially included in the study. All of them were informed in detail about the experiment, and gave their written consent. Two subjects were excluded ‘a posteriori’ from the analysis, one of them due to a previously undocumented partial hearing loss in the range of the carrier tone, and another due to a highamplitude muscle artifact. Therefore, only 10 subjects were included in the final analysis. The protocol was approved by the institutional ethics committee.
Thirteen EEG channels were recorded in monopolar montage referred to both earlobes (F3, Fz, F4, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, according to the 10– 20 International system (Jasper, 1958)), using a commercial electrode cap (Electro-CAP) and Bio-Logic amplifiers. The signal was amplified £20 000; filtered 1– 300 Hz, digitized at 750 Hz and stored in a PC for offline analysis. A minimum of 500 sweeps were recorded per subject, synchronized with the beginning of the sound. The length of the sweep was 2.09 s, with a pre-stimulus period of 0.16 s. The auditory stimuli were designed in the Matlab environment (Mathworks, USA), and delivered through the STIM module of the Neuroscan system (NeuroSoft, El Paso, USA). Bilateral intracanalicular earphones were used. The stimulus consisted of a 85 dB HL 1200 Hz tone modulated in amplitude by a linear chirp, a sinusoid whose frequency increased linearly from 1 to 120 Hz (see Fig. 1). The total length of the sound was 1.61 s. Some additional tests were carried out in a small subgroup of subjects to discard distortions in the response caused by the parameters of the modulation rate. In two of them, longer and shorter sounds (with faster and slower increase of the modulation rate, respectively) were tested and compared with the ‘standard’ responses. Also, in another two subjects, a decreasing-rate chirp modulation (of the same length and frequency range) was used. Finally, in two additional subjects the responses were recorded both in awake and sleep states, using the standard stimulation sound. Three different mathematical approaches were applied to the recorded signal in each experiment. 2.1. Time-frequency transform of the average All sweeps from each subject and condition were averaged, in order to obtain the chirp-evoked potentials. Steady-state potentials are usually studied by means of Fourier transforms. The fast Fourier transform (FFT) gives an accurate estimation of the power spectrum of a stationary signal. However, the responses evoked by a variably modulated tone should not be stationary by definition. Time-frequency transforms can give an estimation of the time-varying energy of a signal in each frequency band (Cohen, 1989). A time-frequency analysis (Gabor transform) was applied to the averaged signal (chirp-evoked potential) from each subject (Gurtubay et al., 2001). The results from all individual subjects were also grand-averaged and displayed using a 3-dimensional (3D) color graph. The same procedure was applied to the modulation of the tone (chirp), in order to assure a constant energy content through time and frequency. 2.2. Average of single-sweep time-frequency transforms Time-frequency transforms are typically used to represent energy values, which are always positive despite
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
701
Fig. 1. Illustration of the procedure followed to build the stimulus. A 1200 Hz tone was modulated in amplitude with a linear chirp, a sinusoid whose frequency increased linearly from 1 to 120 Hz.
the phase. Therefore, the average of single-sweep timefrequency transforms yields an estimation of the energy per sweep taking only into account amplitude changes, and not inter-sweep phase interactions (Tallon-Baudry et al., 1996; Gurtubay et al., 2001). An increase in this graph indicates then that there is an amplitude increase in a frequency at a time point, which is constant in most sweeps. As in the previous analysis, the results were displayed for each individual subject in a 3D colored graph. Also, the average of the transforms was normalized (divided) by the mean energy for each frequency during the pre-stimulus period.
2.3. Inter-trial coherence Any oscillatory activity at a frequency can be defined by two components: amplitude and phase. The procedure outlined above (average of single-sweep time-frequency transforms) yields an estimation of the inter-trial constancy of amplitude changes. As opposed to this analysis, the intertrial coherence (Makeig et al., 2002), also known as phaselocking factor (Tallon-Baudry et al., 1996), gives an estimation of the inter-trial constancy of the phase distribution (see Appendix A). According to this analysis, we obtain a parameter that ranges from 0 (for purely non phase-locked activity) to 1 (for strictly phase-locked activity). The results were displayed in a 3D graph, as in the previous analysis.
3. Results 3.1. Steady-state responses The average of the responses yielded in all subjects an oscillatory potential with several components (Fig. 2, top left). Although some oscillatory changes were already present in some subjects in partial averages of as few as 30 sweeps, the signal-to noise ratio and the amplitude of the changes increased steadily up to 400– 500 sweeps per subject. An initial response, consisting of two consecutive N1-P2 complexes, was observed about 120 ms after the beginning of the sound (mean latency for the first N1 peak: 120, range: 82 –143 ms; mean latency for the second N1 peak: 280 ms). A small biphasic wave was present before it, corresponding to the transient gamma band response (see below). 300 ms after the sound, oscillations at increasing frequencies were observed. The highest amplitude of these oscillations was recorded about 600 ms after the beginning of the sound, what corresponded to a modulation frequency of 45 Hz in the stimulus. The maximum amplitude of the oscillatory response ranged between 2 and 7 mV. A decrease in amplitude was observed in most cases between 900 and 1100 ms, with a late increase between this time and the end of the stimulus (1600 ms). At the end of the stimulus, an offset potential was observed (mean latencies of 56 ms for the negative (N10 ) and 145 ms for the positive peak (P20 ), calculated after the end of the stimulus). A slow positive wave (1 Hz) overlapped the initial part of the oscillatory component.
702
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
Fig. 2. (Left) Grand average of the potential evoked by the chirp-modulated tone (top), and grand average of the time-frequency transforms of the potentials from each subject (bottom), in Fz. The color scale represents absolute energy values. Time 0 corresponds to the beginning of the stimulus. (Top right) Topographical representation of the energy contained in two time-frequency areas, representative of the two main components in the response. (Bottom right) Time-frequency energy distribution of the modulation signal (‘chirp’). Energy values are constant through the diagonal band that increases linearly in frequency along time. The color scale represents relative values.
3.2. Oscillatory activity: time-frequency transform of the evoked potential The results were highly consistent in the whole group of 10 subjects, both in absolute energy values and in energy distribution. The main finding in the plots was a diagonal band of energy, that represented the oscillatory response at the increasing frequency of modulation (Fig. 2, bottom left). This response began approximately at a stimulation frequency of 20 Hz and lasted till the end of the stimulus (120 Hz). The maximum energy was at frequencies around 45 Hz (mean: 45.9, range: 43 –50). In most subjects, two components could be observed in this response, one between 20 and 65 Hz, with higher energy, and a second component between 80 and 120 Hz. The energy content in intermediate frequencies (65 – 80 Hz) was usually much lower. The topography of the lower component (20 –65 Hz) was markedly frontal mesial in the initial montage referred to both earlobes; after computing laplacian derivatives, the maximum energy shifted to both temporal regions. In the parietal and posterior temporal electrodes, the maximum relative amplitude of the response was at 60 Hz, although with lower energy than in frontal leads. The topography of the second component was more diffuse, although it also had frontal preponderance. In half of the subjects, and in the grand average, a second diagonal was observed (with much lower
energy values), which corresponded to a response at twice the stimulation frequency. A small energy peak around 45 Hz was also observed shortly (30 – 35 ms) after the beginning of the stimulus. The frequency of this peak was similar to that of the maximum energy. This peak corresponded to the transient gamma activity evoked by the initial tone (Pantev et al., 1991). A linear representation of the energy changes in representative frequencies of the first and second components of the oscillatory response is shown in Fig. 3; the 40 Hz transient response can be observed in the 45 Hz plot (arrow). Fig. 4 shows the global energy content of the stimulus and the averaged response; the two oscillatory components can be clearly observed in the latter. 3.3. Average of single-sweep time-frequency transforms: amplitude changes The main finding in all subjects was a progressive decrease in energy for higher frequencies during the whole period analyzed, due at least in part to the low-pass filter effect of the skull. Stimulus-related changes were minimal. A ‘diagonal’ band corresponding to the modulation frequency was observed only in two of the subjects (one of them shown in Fig. 5; maximum increase of 60% over baseline). In another 3 subjects, a small increase in energy was observed around 40 Hz only after the normalization of
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
703
Fig. 3. Linear representation of energy changes in Fz along time at two different frequencies, representative of the two main components in the oscillatory response. The dotted line represents energy changes in the modulation signal (chirp), in relative values (right scale). The vertical arrow indicates the 40 Hz transient response. The main peak in both plots matches the frequency of stimulation; the smaller peak observed earlier is a response at twice the frequency of stimulation. Time 0 corresponds to the beginning of the stimulus.
the graph with the pre-stimulus period, at the same time point observed in the time-frequency plot of the averaged potential. Also, a 50 Hz band (mains frequency) was observed in some plots, despite the effort to keep low impedances. Although interindividual variation in absolute energy values was high (as in any non-averaged EEG recording), the proportion of energy in different frequencies was similar.
3.5. Effects of the chirp length and direction The results obtained with larger sweeps (with a 2.5 times slower increase rate of the modulation frequency) were comparable to the results obtained with our ‘standard’ chirp. The increase in the signal-to-noise ratio was minimal,
3.4. Inter-trial coherence/phase-locking factor The results were in all subjects very similar in general shape to those observed in the time-frequency transform of the average. A marked diagonal band was observed in all cases, representing the response to the modulation frequency. Two peaks were observed around 45 and 90 Hz (the same frequencies as in the previous graph). The distribution of the phase-locking factor throughout frequencies matched the energy changes in the TF transform of the average. Also, a small 40 Hz peak corresponding to the 40 Hz transient could be observed in many subjects. The phase-locking values (. 0.4 in all subjects) were similar for the two components of the response (30 – 65 Hz and 80 – 120 Hz). These values were highly significant according to the statistical procedure described by Feige et al. (2000). Fig. 5 includes an example of the 3 mathematical procedures applied to the same data from one representative subject.
Fig. 4. (A) Estimate of the energy spectrum of the modulation signal in the range studied, obtained by adding all energy values at each frequency through the period analyzed. Values are given in relative units. (B) Estimate of the energy spectrum of the grand-averaged response in the range studied, obtained by adding all energy values at each frequency through the period analyzed. Two peaks, corresponding to the main components of the oscillatory response (45 and 95 Hz), can be observed.
704
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
Fig. 5. Time-frequency plots of the 3 mathematical procedures applied to the data from a single representative subject. (Top left) Energy distribution of the averaged response. The color scale represents absolute energy values. (Top right) Phase-locking factor from the same subject. Values can range from 0 to 1 (see text). (Bottom left) Average of single-sweep time-frequency transforms from the same subject. The color scale represents absolute energy values. (Bottom right) Normalization with baseline energy of the bottom left graph. The color scale indicates percentual relative increases or decreases at each frequency compared with the pre-stimulus period. A small band of energy increase can be observed more clearly.
Fig. 6. Responses to an increasing-rate (left) and a decreasing-rate (right) modulated tone in a single subject. The averaged potential (top) and its timefrequency energy distribution (bottom) are plotted at the same scale in both conditions. The energy distribution in frequency in both TF plots is nearly identical, although inverted in time.
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
705
Fig. 7. Responses to a chirp-modulated tone in a single subject during wakefulness (left) and stage 2 sleep (right). The averaged potential (top) and its timefrequency energy distribution (bottom) are plotted at the same scale in both conditions. A description of the changes found during sleep can be found in the main text.
Fig. 8. Averaged responses (top), time-frequency energy distribution (middle), and phase-locking factor (PLF, bottom) obtained in one subject after averaging different numbers of sweeps. Please note that although PLF values apparently decrease when more sweeps are included, the statistical significance increases steadily with an inverse exponential relationship with the number of sweeps. Thus, P values in the low gamma range are hardly significant in the first plots (P , 0:05), and highly significant in the last ones (P , 0:00001). The recording time for each plot is indicated below them.
706
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
despite a much longer duration of the test. Some differences in relative energy values were present in the lower part of the frequency range studied (around 20 –25 Hz) due to the slower increase of the modulation rate, but the absolute maximal values and the distribution of the energy changes in the 30 –120 Hz range were similar. With shorter sweeps (two times faster increase rate of the modulation frequency), the results were much worse in terms of signal-to noise ratio. In some subjects, instead of a diagonal band only isolated peaks of response could be identified in the TF energy distribution. There were no significant differences in the frequency distribution of the energy peaks between the oscillatory responses obtained with an increasing-rate chirp and a decreasing-rate chirp modulation (see Fig. 6), with the exception of a larger 40 Hz transient with the decreasingchirp stimulation. In particular, the frequency of maximal response in both components (around 45 and 90 Hz) in each subject were the same. 3.6. Effects of sleep In the two subjects studied, the amplitude of the responses was smaller during sleep (stage 2), for both components of the response. The decrease in energy was higher for the 40 Hz component of the response (1/6 of awake values) than for the 80 Hz component (1/4). The maximum amplitude of the response was at a slightly lower frequency during the sleep state (36 vs. 41 Hz). Fig. 7 compares the results obtained during sleep and wakefulness in one subject.
4. Discussion 4.1. Chirp-evoked potentials as a new test Our results have shown that it is possible to analyze the oscillatory response to different frequencies of stimulation in a simultaneous way using the time-frequency transform of the potential evoked by a chirp-modulated tone (a tone modulated in amplitude at increasing frequencies). With this technique, the frequency at which the maximum response is obtained, as well as the range of frequencies in which there is an oscillatory response, can be easily determined. The results obtained are comparable to those described in the literature using much longer procedures. A maximum of 500 sweeps of 2.09 s of length (17 min) were required for a good definition of the energy changes in all subjects (see Fig. 8). At least 4 components could be observed in the response. First, an ‘onset’ complex was observed, consisting of the transient 40 Hz response and two long-latency auditory evoked potentials evoked by the two first ‘beats’ (modulation cycles) of the sound. The longer latency of the first long-latency auditory evoked potential and
the transient 40 Hz response compared with other studies can be explained by the modulation of the tone, that caused a very slow rising slope in the first cycle (thus delaying the reaching of the threshold intensity). An ‘offset’ response (Noda et al., 1998) could be also observed in the averaged potential, without a clear correlation with frequency changes in the range studied. Finally, two clear components were observed in the oscillatory response to the frequency of modulation. The first component, with the highest energy, covered the range between 30 and 65 Hz in most subjects. The second component was between 80 and 120 Hz. The origin of these two components is probably very different, according to the literature. The low gamma response (30 – 65 Hz) had a maximum peak around 45 Hz. The frequency of this peak did not differ from the frequency of the transient gamma activity found in a previous study by our group (Gurtubay et al., 2001). This response corresponds to the initial description of the steady-state responses (Galambos et al., 1981). Several studies have located the origin of this response in the auditory cortex (Romani et al., 1982; Makela and Hari, 1987; Makela et al., 1990), but at a location different to the origin of the 40 Hz transient (Pantev et al., 1993). Other studies suggest, however, that there may also be some contribution of brain-stem generators to this response (Hori et al., 1993; Tsuzuku, 1993). The high gamma responses (80 – 120 Hz), on the contrary, are probably mainly originated in the brain-stem (Herdman et al., 2002). We found a maximal response in the low gamma range over Fz and F3, as described in other studies using the same earlobe references (Pastor et al., 2002). The maximum response in the high gamma range had also a frontal predominance. Even though the number of channels used in the present study can not yield a detailed topography, our data are concordant with the literature, suggesting that the responses obtained are equivalent to those described with equivalent constantly-modulated tones. The fact that oscillatory responses at the frequency of stimulation can be obtained with stimuli delivered at variable rates makes the term ‘amplitude modulationfollowing responses (AMFR)’ (Levi et al., 1993) preferable to the term ‘steady-state responses,’ more widely used. The results obtained in the sleep state are fully concordant with previously published data on amplitude changes, thus supporting the validity of our technique to assess the size of the responses at different stimulation frequencies. 4.2. Amplitude and phase Our results have confirmed the known fact that our brain responds to an amplitude-modulated tone with oscillations at the frequency of modulation, within a frequency range. This response is clearly higher for frequencies slightly higher than 40 Hz. A higher amplitude in an average can be due either to higher amplitude in the individual responses, to higher phase constancy in the individual responses, or to
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
a combination of both. In the initial description of the 40 Hz steady-state response, Galambos et al. (Galambos et al., 1981) suggested that this response might be due to a superposition of the middle latency responses evoked by each of the clicks in the series. The results obtained in the phase-locking index clearly indicate that there is a higher constancy in the phase at 40 Hz than at other frequencies of stimulation, thus indicating that phase adjustment is responsible for most of this phenomenon. However, a higher amplitude in the individual response at 40 Hz was also suggested by the small peak in the average of the individual time-frequency transforms, so a higher amplitude in the individual responses to a stimulation at 40 Hz may also be involved. Although the higher signal-to-noise ratio generated by an increment in amplitude might cause an apparent increased phase-synchronization in the phase-locking index plot, the fact that this amplitude increment was only observed in some subjects, and its small size suggests that the findings of phase-synchronization are real. Besides, a higher amplitude in an EEG wave may be caused by higher phase synchronization between the neuronal groups located in the nearby regions. The differences in energy observed in the Gabor transform of the averaged potentials between the low gamma and the high gamma component of the oscillatory response were small or even absent both in the normalized average of the individual Gabor transforms and in the inter-trial coherence plot. The inter-trial coherence plot only represents phase phenomena, without influence of amplitude changes. The average of the Gabor transforms only includes amplitude information, but the normalization corrects for the baseline energy for each frequency, thus giving relative values. This might indicate that the higher energy observed in the low gamma component might be at least partially due to the low-pass filter effect of the skull (Pfurtscheller and Cooper, 1975).
707
synaptic activity at this frequency of stimulation (Pastor et al., 2002). Also, a marked cerebellar activation was also observed only at this frequency. The authors suggested that the cerebellum might be ‘braking’ the extension of the ‘resonant’ cortical activity at this frequency. Our results also suggest that the maximal 40 Hz response may be related to an increased synchronization through the individual responses at this stimulation rate. Why the auditory pathway tends to synchronize better at 40 Hz than at other frequencies? One possible answer is because 40 Hz is one of its usual ‘working’ frequencies. Gamma oscillations may be involved in perceptual binding, as shown in studies with visual or auditory stimulation (Tallon-Baudry et al., 1996; Gurtubay et al., 2001). Some authors have proposed that alterations in the thalamo-cortical circuits involved in the generation of rhythmic gamma activities might be involved in the origin of some neurological diseases (Llinas et al., 1999; McCormick, 1999). There are several descriptions of altered responses to steady-state auditory or visual stimulation in neurological or psychiatric diseases (Butler et al., 2001). A displacement towards lower frequencies of the maximum response to auditory steady-state stimuli has been described in Parkinson’s disease (Pastor and Artieda, 1996). The variability observed in the maximum energy peak was minimal in our study. Therefore, our method provides a fast and reliable test to determine the range of frequencies and the frequency of maximal response to auditory rhythmic stimulation that can be used for the comparison between normal subjects and different pathologies.
Acknowledgements This work has been supported by grant 030043 from the Fondo de Investigaciones Sanitarias, Spain.
4.3. The meaning of the 40 Hz response There are several theories on the origin of the 40 Hz response. As commented above, it was initially thought that it was due to a superposition of the auditory middle-latency evoked potentials (Galambos et al., 1981; Hari et al., 1989). Some recent studies using a multiple generator model support this idea (Gutschalk et al., 1999). However, the persistence of the steady-state responses after the end of the stimulus indicate that resonance phenomena (Bas¸ar et al., 1987) might contribute to this response (Santarelli et al., 1995). Our results do not support either the superposition theory, as we obtained the same results (maximal responses around 40 Hz) using a variable modulation, that implies a progressively shorter cycle hampering the superposition of waves. A recent study using PET showed that there is a higher activation of the auditory cortex during a 40 Hz steady-state stimulation than with other frequencies, indicating a higher
Appendix A The degree of phase resetting (i.e. phase consistency or phase locking) of EEG activity in single trials can be estimated with the procedure proposed by Tallon-Baudry et al. (1996). The method yields a parameter named phaselocking factor, also known as inter-trial coherence (Makeig, 2002), that gives an estimation of the inter-trial constancy of the phase distributions. For each trial, we compute its convolution with a complex wavelet centered in the frequencies of interest. Here we have used a Morlet wavelet
wtðtf Þ ¼ e
2
t2 2s2t
ej2pft
708
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709
with
st ¼
1 2psf
The wavelet family is characterized the constant ratio: Nh0 ¼ st ·f ; we have used Nh0 ¼ 10: Next, we extract the phase of this convolution fðt; f ; nÞ for every time instant t, frequency f and trial n ¼ ½1; …; N: The inter-trial coherence value is then defined for every time and frequency bin as the average across trials. ccomplex ðt; f Þ ¼
N 1 X ejfðt;f ;nÞ N n¼1
The amplitude c of the resulting average represents a measure of the uniformity of the distribution of phases, i.e. how consistent the phase resettings across trials are. In order to provide a statistical significance level to the inter-trial coherence values, we have used the approximation proposed by Feige et al. (2000). Their work was focused on phase differences; our purpose is similar, as we try to measure how far from a uniform distribution the calculated phase distributions are. According to this test, the statistical significance of a c inter-trial coherence value determined from N trials can be calculated as P ¼ e2Nc
2
where P, for a given value of ITC c calculated from N trials, describes the probability to obtain an equal or larger ITC value by chance.
References Bas¸ar E, Rosen B, Bas¸ar-Eroglu C, Greitschus F. The associations between 40 Hz-EEG and the middle latency response of the auditory evoked potential. Int J Neurosci 1987;33:103–17. Bas¸ar E, Bas¸ar-Eroglu C, Parnefjord R, Rahn E, Schurmann M. Evoked potentials: ensembles of brain induced rhythmicities in the alpha, theta and gamma ranges. In: Bas¸ar E, Bullock TH, editors. Induced rhythms in the brain. Brain dynamic series, Boston, MA: Birkhauser; 1992. p. 155 –81. Butler PD, Schechter I, Zemon V, Schwartz SG, Greenstein VC, Gordon J, Schroeder CE, Javitt DC. Dysfunction of early-stage visual processing in schizophrenia. Am J Psychiatry 2001;158:1126–33. Cohen L. Time-frequency distributions. A review. Proc IEEE 1989;77: 941– 81. Engel AK, Konig P, Kreiter AK, Schillen TB, Singer W. Temporal coding in the visual cortex: new vistas on integration in the nervous system. Trends Neurosci 1992;15:218 –26. Feige B, Aertsen A, Kristeva-Feige R. Dynamic synchronization between multiple cortical motor areas and muscle activity in phasic voluntary movements. J Neurophysiol 2000;84:2622 –9. Galambos R, Makeig S, Talmachoff PJ. A 40-Hz auditory potential recorded from the human scalp. Proc Natl Acad Sci USA 1981;78: 2643–7. Gurtubay IG, Alegre M, Labarga A, Malanda A, Iriarte J, Artieda J. Gamma band activity in an auditory oddball paradigm studied with the wavelet transform. Clin Neurophysiol 2001;112:1219–28.
Gutschalk A, Mase R, Roth R, Ille N, Rupp A, Hahnel S, Picton TW, Scherg M. Deconvolution of 40 Hz steady-state fields reveals two overlapping source activities of the human auditory cortex. Clin Neurophysiol 1999;110:856–68. Hari R, Hamalainen M, Joutsiniemi SL. Neuromagnetic steady-state responses to auditory stimuli. J Acoust Soc Am 1989;86:1033 –9. Herdman AT, Lins O, Van Roon P, Stapells DR, Scherg M, Picton TW. Intracerebral sources of human auditory steady-state responses. Brain Topogr 2002;15:69– 86. Hori A, Yasuhara A, Naito H, Yasuhara M. Steady-state auditory evoked potentials (SSAEPs) in the rabbit. Contribution of the inferior colliculus. Electroenceph clin Neurophysiol 1993;88:229–36. Jasper HH. Report of the Committee on Methods of Clinical Examination in Electroencephalography. Electroenceph clin Neurophysiol 1958;10: 370 –5. John MS, Picton TW. Human auditory steady-state responses to amplitudemodulated tones: phase and latency measurements. Hear Res 2000;141: 57 –79. Levi EC, Folsom RC, Dobie RA. Amplitude-modulation following response (AMFR): effects of modulation rate, carrier frequency, age, and state. Hear Res 1993;68:42– 52. Lins OG, Picton PE, Picton TW, Champagne SC, Durieux-Smith A. Auditory steady-state responses to tones amplitude-modulated at 80– 110 Hz. J Acoust Soc Am 1995;97:3051 –63. Llinas RR, Ribary U, Jeanmonod D, Kronberg E, Mitra PP. Thalamocortical dysrhythmia: a neurological and neuropsychiatric syndrome characterized by magnetoencephalography. Proc Natl Acad Sci USA 1999;96:15222–7. Makeig S, Westerfield M, Jung TP, Enghoff S, Townsend J, Courchesne E, Sejnowski TJ. Dynamic brain sources of visual evoked responses. Science 2002;295:690– 4. Makela JP, Hari R. Evidence for cortical origin of the 40 Hz auditory evoked response in man. Electroenceph clin Neurophysiol 1987;66:539–46. Makela JP, Karmos G, Molnar M, Csepe V, Winkler I. Steady-state responses from the cat auditory cortex. Hear Res 1990;45:41– 50. McCormick DA. Are thalamocortical rhythms the Rosetta Stone of a subset of neurological disorders? Nat Med 1999;5:1349–51. Noda K, Tonoike M, Doi K, Koizuka I, Yamaguchi M, Seo R, Matsumoto N, Noiri T, Takeda N, Kubo T. Auditory evoked off-response: its source distribution is different from that of on-response. Neuroreport 1998;9: 2621–5. Pantev C, Makeig S, Hoke M, Galambos R, Hampson S, Gallen C. Human auditory evoked gamma-band magnetic fields. Proc Natl Acad Sci USA 1991;88:8996–9000. Pantev C, Elbert T, Makeig S, Hampson S, Eulitz C, Hoke M. Relationship of transient and steady-state auditory evoked fields. Electroenceph clin Neurophysiol 1993;88:389–96. Pastor MA, Artieda J. Cortical oscillatory behaviour to auditory repetitive stimulation in Parkinson’s disease. Mov Disord 1996;11(Suppl):128. Pastor MA, Artieda J, Arbizu J, Marti-Climent JM, Penuelas I, Masdeu JC. Activation of human cerebral and cerebellar cortex by auditory stimulation at 40 Hz. J Neurosci 2002;22:10501–6. Penny W, Kiebel S, Kilner J, Rugg M. Event-related brain dynamics. Trends Neurosci 2002;25:287–9. Pfurtscheller G, Cooper R. Frequency dependence of the transmission of the EEG from cortex to scalp. Electroenceph clin Neurophysiol 1975; 38:93–6. Pfurtscheller G, Lopes da Silva FH. Functional meaning of event-related desynchronization (ERD) and synchronization (ERS). In: Pfurtscheller G, Lopes da Silva FH, editors. Event-related desynchronization. Handbook of electroencephalography and clinical neurophysiology. Revised series, Amsterdam: Elsevier; 1999. p. 51 –65. Picton TW, Skinner CR, Champagne SC, Kellett AJ, Maiste AC. Potentials evoked by the sinusoidal modulation of the amplitude or frequency of a tone. J Acoust Soc Am 1987;82:165–78. Regan D. Evoked potentials in basic and clinical research. In: Remond A, editor. EEG informatics. A didactic review and applications of EEG
J. Artieda et al. / Clinical Neurophysiology 115 (2004) 699–709 data processing. Amsterdam: Elsevier/North Holland Biomedical; 1977. p. 319–46. Rodriguez E, George N, Lachaux JP, Martinerie J, Renault B, Varela FJ. Perception’s shadow: long-distance synchronization of human brain activity. Nature 1999;397:430–3. Romani GL, Williamson SJ, Kaufman L. Tonotopic organization of the human auditory cortex. Science 1982;216:1339–40. Santarelli R, Maurizi M, Conti G, Ottaviani F, Paludetti G, Pettorossi VE. Generation of human auditory steady-state responses (SSRs). II. Addition of responses to individual stimuli. Hear Res 1995;83: 9–18.
709
Singer W. Synchronization of cortical activity and its putative role in information processing and learning. Annu Rev Physiol 1993;55: 349– 74. Stapells DR, Linden D, Suffield JB, Hamel G, Picton TW. Human auditory steady state potentials. Ear Hear 1984;5:105–13. Tallon-Baudry C, Bertrand O, Delpuech C, Pernier J. Stimulus specificity of phase-locked and non-phase-locked 40 Hz visual responses in human. J Neurosci 1996;16:4240 –9. Tsuzuku T. 40-Hz steady state response in awake cats after bilateral chronic lesions in auditory cortices or inferior colliculi. Auris Nasus Larynx 1993;20:263–74.