- Email: [email protected]
Wavelet-based estimation of EEG coherence during Chinese Stroop task
Computers in Biology and Medicine 36 (2006) 1303 – 1315 www.intl.elsevierhealth.com/journals/cobm
Wavelet-based estimation of EEG coherence during Chinese Stroop task Xiaofeng Liu, Huan Qi, Supin Wang, Mingxi Wan∗ The Key Laboratory of Biomedical Information Engineering of Ministry of Education, and Department of Biomedical Engineering, School of Life Science and Technology, Xi’an Jiaotong University, Xi’an 710049, PR China Received 16 May 2005; accepted 11 August 2005
Abstract Wavelet-based estimation of instantaneous EEG coherence was applied to investigate the synchronization of different brain regions whilst 10 subjects performing Stroop task presented in Chinese. In contrast to coherence based on Fourier transform, wavelet-based coherence, which does not depend on the stationarity of signals, applies an adaptive window to the frequency of the signal and has a more accurate time-frequency resolution. In the present study, a greater negativity for the incongruent situation than congruent situation appeared from 350 to 600 ms post-stimulus onset over frontal, central, and parietal regions, and significantly higher EEG coherences for the incongruent situation than congruent situation were observed over frontal, parietal, and frontoparietal regions from 100 to 400 ms at 1 (13–18 Hz) frequency band, which was found to be sensitive in the discrimination between congruent and incongruent situations. The findings in the present study may indicate that functional synchronization as indexed by EEG coherence at 1 frequency band is enhanced at the earlier stage while processing the conflicting information from the incongruent stimulus, and that 1 frequency band is close related to interactions of brain areas in the selected attention task. 䉷 2005 Elsevier Ltd. All rights reserved. Keywords: EEG; Coherence; Wavelet; Chinese Stroop task; Functional synchronization
∗ Corresponding author. Tel.: +86 29 82667924; fax: +86 29 82668668.
E-mail address: [email protected] (M. Wan). 0010-4825/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiomed.2005.08.002
1304
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1. Introduction The coherence of scalp recorded EEG is a frequency domain measure of similarity between EEG signals and gives a measure of the degree of functional cooperation between neuronal systems underlying the generation of those signals [1]. The EEG coherence has proven itself a promising tool for investigating the synchronization of the cortical activity during various cognitive processes [2–6]. The Stroop task [7], in which subjects are required to name the color of color-word printed in congruent (e.g., RED in red) or incongruent (e.g., RED in blue) ink as quickly as possible while ignoring the meaning of word, involves selective attention, language processing, color-naming processes. Therefore, the investigation of the synchronization of functional networks during Stroop task by means of coherence may be helpful to understand the neurophysiological basis of high-level cognitive activities of normal humans, and even to provide clinical assessments for dysfunctions or disorders in patients. In numerous studies on the Stroop task, event-related potential (ERP) is often employed to explore the temporal course of the related neural processes because of its capability to record brain activity with a millisecond level temporal resolution. Recently, the electrophysiological correlates of the Stroop task with three different response modalities (Overt Verbal, Covert Verbal, and Manual) were investigated [8]. It was found that a negativity, the negative-going ERP waveform, was greater for the incongruent situation than congruent situation at 350–500 ms post-stimulus onset (peaking at 410 ms) over midline sites, which were interpreted as evidences of conflict processing and resolution in the anterior cingulate. In addition, a greater positive shift of ERP for the incongruent situation than congruent situation was observed at 500–800 ms over the left temporo-parietal regions and negativity over the anterior frontal region. Another ERP study [9] of Stroop effect also reported a greater negativity for the incongruent situation than congruent situation over the fronto-central region beginning at about 500 ms and a greater positive shift over the left temporo-parietal regions beginning at approximately 650 ms. In the previous studies, the neural activity associated with Stroop task over the independent electrode site was mainly researched. Consequently, the temporal course and distribution of the neural activity were analyzed and reported in detail, but not the functional interaction of the different areas. Nowadays, it is widely held that different regions of the brain must cooperate with each other to support integration of sensory information, sensory-motor coordination and many other functions that are critical for learning, memory, information processing, perception and the behavior of organisms [10]. West and Bell [11] observed an increased alpha-1 (8–10 Hz) within the prefrontal and parietal areas when they examined EEG power for different frequency bands over the frontal and parietal regions in a Stroop study, and assumed an interaction between prefrontal and parietal regions. Schack et al. [12] firstly addressed the interaction processes of different brain regions in the Stroop task in relation to the time-evolution of the cognitive process by means of instantaneous coherence analysis based on Fourier transform (FT). Their research revealed that the frequency band of 13–20 Hz was sensitive to the discrimination between the congruent and incongruent situations, and that higher coherences were observed within the left frontal and left parietal areas, as well as between them for the incongruent situation in comparison with the congruent situation. Unfortunately, vocal response modality in Stroop task was employed, and thus, the facial muscle movements underlying speech production could introduce a significant artifact into EEG. In the present study, the neural correlates of Chinese Stroop task are addressed by using high-density EEG recordings. Then, the properties of coherence patterns of different brain areas in different conditions
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1305
during Stroop task are examined by means of wavelet-based coherence that has been often applied to the estimation of coherence among non-stationary signals [13–16]. Wavelet-based coherence has a more accurate time-frequency resolution than the coherence based on the windowed short-term Fourier transform, in that the size of the window is adapted to the frequency of the signal in wavelet analysis, while it is fixed in short-term Fourier transform; thus, it provides the time-varying power spectrum as well as the phase spectrum.
2. Materials and methods 2.1. Subjects Ten healthy native Chinese (aged 21–30, mean = 23.4 years, five males and five females) participated in this study with informed consent. All subjects were right handed according to a self-report, and free from neurological or psychiatric illness. All had normal or corrected to normal visual acuity, and normal color vision. 2.2. Stimuli and procedure Stimuli (Chinese equivalents of Red, Green and Blue) were printed in the color red, or green, or blue and presented in the center of the screen against a black background. The stimulus subtended approximately 4.9◦ visual angle horizontally and 1.4◦ vertically. Subjects were seated inside a dimly lit room at a distance of approximately 70 cm from the screen. The experiment included a color-to-key acquisition phase, a practice phase, and a test phase. In the color-to-key acquisition, subjects performed a block of 60 trials (XXX was printed in each of the three colors appeared 20 times) to establish a strong mapping between the stimulus color and corresponding key. Also, a practice task of 36 trials (color-word stimulus) was completed to acquaint the subject with the task. Each subject participated two blocks of 90 trials (equivalent trials for congruent and incongruent). Each trial of a block was pseudo-randomized presented. Subjects were instructed to pay attention to the word presented on the screen and response to ink color of the words as quickly and accurately as possible by pressing the appropriate key. Stimuli duration was 100 ms, and stimulus onset asynchrony (SOA) varied randomly between 1600 and 1800 ms to minimize the contribution of expectation effects on the stimulus-locked EEG. The block presentation order was counterbalanced across subjects. 2.3. EEG recording EEG activity was continuously recorded from the scalp using EEG amplifier (SynAmps, NeuroScan Inc., USA) with a 32-channel electrode-cap (Quik-Cap). All electrodes were referenced to linked electrode placed on the left and right mastoids (International 10/20-Systems). EEG channels were amplified using a band-pass of 0.1–45 Hz, sampling frequency at 250 Hz, impedence below 5 K, and a gain of 12 500. Electrooculogram (EOG) activity was recorded from four electrodes (two placed above and below the left eye; two placed at external canthus of each eye) and used as a guide to removal of eye movement artifacts and blinks. ERP analysis epochs were extracted offline and included 100 ms pre-stimulus and
1306
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
900 ms post-stimulus. Any trial on which EOG or EEG amplitude exceeded ±100 V was rejected as artifact. Only EEGs of trials associated with a correct response and not contaminated with artifacts were averaged according to stimulus type. 2.4. Wavelet coherence The continuous wavelet transform of a square integrable complex-valued signal x(t) is defined as Wx (, f ) =
+∞ −∞
x(t)∗,f (t) dt.
(1)
Morlet wavelet is simple and well suited for spectral estimations. It is defined for frequency f and time by (u − )2 , (2) ,f (u) = f exp i2f (u − ) exp − 2
,f (u) is simply the product of a sinusoidal wave at frequency f, with a Gaussian function centered at time with a standard deviation proportional to the inverse of f. From the wavelet transforms of two signals x(t) and y(t), the wavelet auto-spectrum of x(t) and/or y(t) and wavelet cross-spectrum between x(t) and y(t) are defined as w Sxx (t, f ) = Wx (, f )Wx∗ (, f ) d, (3) w Sxy (t, f ) =
T
T
Wx (, f )Wy∗ (, f ) d.
(4)
The localized time integration window in Eqs. (3 and 4) T = [t − , t + ] is selected based on the time resolution desired in the resulting coherence map ( can be adapted to the frequency of interest); thus, the coherence map provides a view of localized correlation with respect to both time and frequency. Finally, the wavelet coherence w xy (t, f ) is defined at time t and frequency f by w xy (t, f ) =
w (t, f )| |Sxy
(5)
w (t, f )S w (t, f ) Sxx yy
w (t, f ), a quantity bounded by zero and unity, The magnitude squared coherence (MSC), denoted by Cxy is defined at time t and frequency f by w 2 Cxy (t, f ) = |w xy (t, f )| = w 0 Cxy (t, f ) 1 ∀f .
w (t, f )|2 |Sxy
w (t, f )S w (t, f ) Sxx yy
.
(6a) (6b)
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1307
The average of the wavelet coherence w xy (t, f ) between frequency bands enables a necessary data reduction. As a result, mean band coherence as a function of time is obtained. For a chosen frequency band [fl , fu ] the mean band coherence is computed by fu w (t) = w (7a) xy (t, f ) df , f fl
2 Cfw = |w f| .
(7b)
Further, mean coherence within a chosen time-frequency region [t1 , t2 ] ∗ [fl , fu ] is defined at time t by t2 f u w (t) = w (8a) xy (t, f ) df dt, t,f t1
fl
w 2 = |w Ct,f t,f | .
(8b)
The coherence analysis was performed for adjacent electrode pairs within the frontal, central, temporal, and parietal regions and additionally for frontoparietal electrode pairs, such as F3/P3, F3/Pz, F3/P4, Fz/P3, Fz/Pz, Fz/P4, F4/P3, F4/Pz, F4/P4, and temporoparietal electrode pairs, such as TP7/P3, P3/T7, TP7/T7, TP8/P4, P4/T8, and TP8/T8. 3. Results 3.1. Event-related potentials Fig. 1 showed the grand average ERP elicited by congruent and incongruent trials for correct responses. All trials elicited a late positive complex wave with maximum over the midline centro-parietal region. A greater negativity for the incongruent situation than congruent situation, peaking at around 440 ms, was observed between 350 and 600 ms post-stimulus onset over frontal, central, and parietal regions (see Fig. 1). The current results were consistent with previous ERP studies [8,9]. 3.2. Coherence analysis 3.2.1. Time frequency map of EEG coherence in congruent and incongruent task situations Due to wavelet transform, it is possible to examine the instantaneous EEG coherences at different frequencies by means of time-frequency representations (TFRs). In the present study, wavelet coherence was performed for the electrode pairs Fp1/F3, Fp1/F7, F7/F3, F3/Fz and F7/Fz within the left frontal area, for the electrode pairs Fp2/F4, Fp2/F8, F8/F4, F4/Fz and F8/Fz within the right frontal area, for the electrode pairs TP7/P3, P3/T7, and TP7/T7 within the left temporoparietal area, for the electrode pairs TP8/P4, P4/T8, and TP8/T8 within the right temporoparietal area and for the frontoparietal electrode pairs F3/P3, F3/Pz, F4/P4 and F4/Pz. We estimated wavelet coherence separately for all the single trials, and then averaged the resulting distributions to enhance time-frequency regions bearing similar timefrequency signatures. As a result, higher coherences were found over the electrode pairs within and between the frontal and the parietal regions. Fig. 2 showed the time-frequency map of wavelet coherence of the electrode pair F3/Pz in the congruent and incongruent situations and the difference between them
1308
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 F3
F1
0
0
4
4 0
400 ms
800
0
C3
400 ms
800
C1
0
0
4
4 0
400 ms
0
800
CZ
400 ms
800
PZ
0
0
4
4 0
400 ms
800
0
T7 0
4
4 400 ms
800
T8
0
0
400 ms
800
0
400 ms
800
Fig. 1. Grand average ERPs over 10 subjects at selected electrode sites F3, F4, C3, C4, Cz, Pz, T7, and T8 for the congruent (bold real line) and incongruent (thin dotted line) conditions.
(incongruent minus congruent situation), averaged for all single trials and all 10 subjects. To make it clear and enhance the comparison, we selected out the values higher than the threshold (herein, 0.5) of the differences between the incongruent and congruent situations, which were shown in the lower right panel of Fig. 2. In Fig. 2, higher coherences were observed at frequency band of beta (13–30 Hz) and gamma (> 30 Hz) for both incongruent and congruent situations, and higher difference between the incongruent and congruent situations appeared at the beta frequency band before around 400 ms post-stimulus onset. In order to objectively assess the degree of coherence between different electrode pairs, the mean wavelet coherence was calculated for all 10 subjects and both task conditions. Generally, high coherences appeared during the whole task for higher frequency bands, beta (13–30 Hz) and gamma (> 30 Hz) frequency bands, for the two task situations, and mainly distributed within frontal, parietal, frontal-central, and fronto-parietal regions. According to the TFRs of the coherence difference (incongruent minus congruent situation) within and between the frontal, central, parietal, and temporal regions, salient coherence difference appeared in the time-frequency region, approximately at [100–400 ms]∗[10–20 Hz], for electrode pairs within and between frontal and parietal regions, such as Fp1/F3, Fp1/F7, F7/F3, F3/Fz, Fp2/F4, Fp2/F8, F8/F4, F4/Fz, P3/Pz, P4/Pz, F3/P3, F3/Pz, F4/P4, and F4/Pz, namely, the highest differences between congruent and incongruent situations appeared approximately within this time-frequency region. To further assess the sensitivity of different frequency bands for the discrimination of the two task situations, the mean band coherences for standard frequency bands were calculated according to Eq. (7b). The mean time courses of band coherences for the electrode pair F3/Pz were shown in Fig. 3. The largest
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 Congruent
1309
InCongruent 1.0
1
10
10
0.75
0.75
20 Hz
Hz
20 0.50
30 0.25
40 50
0.50 30
0.0 0
200
400 ms
600
0.25
40 50
800
0
Incongruent-Congruent Difference
200
400 ms
600
800
0.0
Selected Difference (> 0.5) 1.0 0.6
10
10 0.75
0.3 20 0.0 30
Hz
Hz
20 0.50 30 -0.3
40
0.25
40 -0.6
50
0
200
400 ms
600
800
50
0
200
400 ms
600
800
0.0
Fig. 2. Time-frequency map of wavelet coherence of the electrode pair F3/Pz averaged for all single trials and all 10 subjects. Upper left panel: Congruent; Upper right panel: Incongruent; Lower left panel: Incongruent minus Congruent difference; Lower right panel: Selected difference between incongruent and congruent situations (only values higher than 0.5 were kept). Color-scale from ‘black’ (minimum) to ‘white’ (maximum).
coherence difference between incongruent and congruent situations appeared at the 1 frequency band (13–18 Hz) for the electrode pair F3/Pz in Fig. 3, so did they for frontal and parietal areas in the left hemisphere. The mean time courses of band coherences averaged over the electrode pairs F7/F3, F7/Fz, F3/Fz, F3/P3, F3/Pz, F7/P3, F7/Pz, and P3/Pz in the left hemisphere were represented in Fig. 4. The coherence difference between incongruent and congruent situations at the 1 frequency band for left hemispherical frontal and parietal areas lasted for longer time than that for the single electrode pair F3/Pz. A similar effect was also found for the coherences averaged over the electrode pairs in the right hemisphere. The mean band coherences averaged over the electrode pairs F8/F4, F8/Fz, F4/Fz, F4/P4, F4/Pz, F8/P4, F8/Pz, and P4/Pz of the right hemisphere were shown in Fig. 5 . On this basis, the frequency band of 13–18 Hz was roughly deemed sensitive to discriminate the congruent and incongruent task situations, and highest differences of coherence between the two situations mainly occurred at around 100–400 ms.
1310
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.6
0.2
0.4
0
0.2
0
200
400
600
-0.2
0
200
ms
(a)
400
600
400
600
400
600
400
600
ms 0.2
0.8
0.1 0.6 0 0.4
0
200
(b)
400
600
-0.1
0
200
ms
ms
1
0.2 0.1
0.8 0 0.6
-0.1 0
200
(c)
400
600
0
200
ms
ms
1 0.1 0.8
0 -0.1
0.6 (d)
0
200
400 ms
600
0
200 ms
Fig. 3. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent situations (gray real line, right panel) for the electrode pair F3/Pz for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz. The thin vertical line denotes the mean reaction time for the congruent situation (483 ms), the thick vertical line denotes the mean reaction time for the incongruent situation (527 ms).
3.2.2. Topographical map of EEG coherence at a sensitive time-frequency region As described in the preceding section, the frequency band of 13–18 Hz had proven to be sensitive for discrimination between the congruent and incongruent task situations within 100–400 ms. The mean coherences within the time-frequency region, [13–18 Hz]∗[100–400 ms], over these electrode pairs within and between frontal, central, temporal, parietal, and occipital regions, were calculated according to Eq. (8b), and then the topography of coherences of the two task situations was obtained and illustrated in Fig. 6. The mean coherence averaged over Fp1/F7, Fp1/F3, F3/F7, F3/Fz, F7/Fz, F3/P3, F3/Pz, F7/P3, and F7/Pz represents the left hemisphere and the mean coherence averaged over Fp2/F8, Fp2/F4, F4/F8, F4/Fz, F8/Fz, F4/P4, F4/Pz, F8/P4, and F8/Pz represents the right hemisphere, analogously. There was no significant difference in coherence between the hemispheres. In addition, coherences for the incongruent situation were significantly stronger than those for the congruent situation in not only the left hemisphere
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.8
0.2
0.6
0
0.4 (a)
0
200
400
600
-0.2
0
200
ms
1311
400
600
400
600
400
600
400
600
ms 0.2
0.8
0.1 0
0.6 (b)
0
200
400
600
-0.1
0
200 ms
ms 0.2
0.8
0.1 0
0.6 (c)
0
200
400
600
-0.1
0
200
ms
ms 0.2
0.8
0.1 0
0.6 0 (d)
200
400 ms
600
-0.1
0
200 ms
Fig. 4. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent situations (gray real line, right panel) for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz averaged over the left hemispherical electrode pairs F7/F3, F7/Fz, F3/Fz, F3/P3, F3/Pz, F7/P3, F7/Pz, P3/Pz, F3/O1, F3/Oz, F7/O1, F7/Oz.
but also the right hemisphere, as shown in Fig. 6. Therefore, the synchronization related to the Stroop effect seems to be symmetrical according to the coherence of the sensitive frequency band 13–18 Hz.
4. Discussion The aforementioned coherence results were achieved with linked mastoid reference. Although, the reference electrode’s influence on EEG coherence was demonstrated in several studies [17–20], the comparison between the congruent and incongruent task situations could be performed on the basis of instantaneous coherence, as Schack and his colleagues argued [12]. EEG coherence analysis, an important tool for studying high-level cognitive processes, enables us to examine the intermediate sub-processes of interaction among different topographic areas and thus gives insights into the investigation of the functional networks cooperation during various cognitive processes.
1312
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.8
0.2
0.6
0
0.4 (a)
0
200
400
600
-0.2
0
200
ms
400
600
400
600
400
600
400
600
ms
0.8 0.2 0.6
0.1
0.4
-0.1
0 0
200
(b)
400
600
0
200 ms
ms
0.8
0.2 0.1 0
0.6 (c)
0
200
400
600
-0.1
0
200
ms
ms 0.2
0.8
0.1 0
0.6
-0.1 0
(d)
200
400 ms
600
0
200 ms
Fig. 5. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent (gray real line, right panel) for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz averaged over the right hemispherical electrode pairs F8/F4, F8/Fz, F4/Fz, F4/P4, F4/Pz, F8/P4, F8/Pz, P4/Pz, F4/O2, F4/Oz, F8/O2, F8/Oz.
Therefore, coherence between bioelectrical signals has been frequently applied to measure the relationship between EEG/MEG signals recorded during cognitive function of normal human [2–5,21] and psychiatric or neurological patients [22–24]. The wavelet coherence is very different from the traditional spectral coherence, in that it describes the correlation of multiscale wavelet components rather than fixed-scale sinusoidal components of the EEG signals. In the present study the wavelet coherence was applied to address the interference during incongruent Stroop task. High coherences appeared during the whole task for frequency band beta (13–30 Hz) and gamma (> 30 Hz) for the two situations in the present study. The coherences within time interval 100–400 ms at the frequency band 13–18 Hz were overall higher for incongruent situations than those for congruent situations over the electrode pairs within the frontal, central, and parietal regions, and additionally for frontoparietal regions. The frequency band beta-1 (13–18 Hz) was found sensitive to the discrimination between the incongruent and congruent situation, which was consistent with the research of Schack and his colleagues [12]. Coherences for the incongruent situation were significantly stronger than those for
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 Congruent
Incongruent
Fp2
Fp1
T7
TP7
FP1 F8
F7
FT7
F3
Fz
F4
FC3
FCz
FC4
C3
Cz
C4
CP3
CPz
CP4
P3
Pz Oz
O2
F8 Fz
F4
FCz
FC4
C2
Cz
C4
CP3
CPz
CP4
P3
Pz
P4
O1
Oz
O2
F3 FT8
FP2
P7
PT7
T8
T7
TP8
TP7
P9
P7
P4
P7 O1
1313
C3
FT8
T8
TP8
P9
Fig. 6. Topographical maps of the mean coherence at frequency band 13–18 and time interval 100–400 ms (weight of line is proportional to magnitude of coherence).
the congruent situation in both the left hemisphere and the right hemisphere. Therefore, it seemed to be plausible that interaction or synchronization as indexed by EEG coherence between the functional brain cortical areas was enhanced when processing the incongruent stimulus, according to the higher coherence for the incongruent situation than congruent situation. In the present study, the Stroop effect seems to be symmetrical according to the synchronization as indexed by coherence, which is distinct to the findings of Schack et al. [12] who observed a left lateralized synchronization phenomenon. This may be attributed to the difference in the processing of the Chinese stimulus and English stimulus, and needs further investigation. Acknowledgements The authors gratefully acknowledge the support from the National Natural Science Foundation of China (Grants 30070212 and 69925101). References [1] P.L. Nunez, EEG coherence measures in medical and cognitive science: a general overview of experimental methods, computer algorithms, and accuracy, in: H. Witte, U. Zwiener, B. Schack, A. Doering (Eds.), Quantitative and Topological EEG and MEG Analysis, Druckhaus Mayer, Jena, 1997, pp. 385–392. [2] H. Petsche, S.C. Etlinger, O. Filz, Brain electrical mechanism of bilingual speech management: an initial investigation, Electroenceph. Clin. Neurophysiol. 86 (1993) 385–394. [3] H. Petsche, Approaches to verbal, visual and musical creativity by EEG coherence analysis, Int. J. Psychophysiol. 24 (1996) 145–159. [4] H. Petsche, S.C. Etlinger, EEG and Thinking, Oesterreichische Akademie der Wissenschaften, Wien, 1998. [5] J. Sarnthein, H. Petsche, P. Rappelsberger, G.L. Shaw, A. von Stein, Synchronization between prefrontal and posterior association cortex during human working memory, Proc. Natl. Acad. Sci. USA 95 (1998) 7092–7096.
1314
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
[6] S. Weiss, H. Mueller, The contribution of EEG coherence to the investigation of language, Brain Lang. 85 (2003) 325–343. [7] J.R. Stroop, Studies of interference on serial verbal reactions, J. Exp. Psychol. 18 (1935) 643–662. [8] M. Liotti, M.G. Woldorff, R. Perez, H.S. Mayberg, An ERP study of the temporal course of the Stroop color-word interference effect, Neuropsychologia 38 (2000) 701–711. [9] R. West, C. Alain, Event-related neural activity associated with the Stroop task, Cognitive Brain Res. 8 (1999) 157–164. [10] W.H.R. Miltner, C. Braun, M. Arnold, H. Witte, E. Taub, Coherence of gamma-band EEG activity as a basis for associative learning, Nature 397 (1999)434–436. [11] R. West, M.A. Bell, Stroop color-word interference and electroencephalogram activation: evidence for age-related decline of the anterior attention system, Neuropsychology 11 (1997) 421–427. [12] B. Schack, A.C. Chen, S. Mescha, H. Witte, Instantaneous EEG coherence analysis during the Stroop task, Clin. Neurophysiol. 110 (1999) 1410–1426. [13] P. Liu, Wavelet spectrum analysis and ocean wind waves, in: E. Foufoula-Georgiou, P. Kumar (Eds.), Wavelets in Geophysics, Academic Press, New York, 1994, pp. 151–166. [14] S. Santoso, E. Powers, R. Bengtson, A. Ouroua, Time-series analysis of nonstationary plasma fluctuations using wavelet transforms, Rev. Sci. Instrum. 68 (1997) 898–901. [15] K. Gurley, T. Kijewski, A. Kareem, First- and higher-order correlation detection using wavelet transforms, J. Eng. Mech. 129 (2) (2003) 188–201. [16] J.P. Lachaux, A. Lutz, D. Rudrauf, D. Cosmelli, M. Le Van Quyen, J. Martinerie, F. Varela, Estimating the time-course of coherence between single-trial brain signals: an introduction to wavelet coherence, Clin. Neurophysiol. 32 (2002) 157–174. [17] C. Andrew, G. Pfurtscheller, Dependence of coherence measurements on EEG derivation type, Med. Biol. Eng. Comput. 34 (1996) 232–238. [18] T.D. Lagerlund, F.W. Sharbrough, N.E. Busacker, K.M. Cicora, Interelectrode coherences from nearest-neighbor and spherical harmonic expansion computation of Laplacian of scalp potential, Electroenceph. Clin. Neurophysiol. 95 (1995) 178–188. [19] P.L. Nunez, R. Srinivasan, R.S. Wijesinghe, A.F. Westdorp, D.M. Tucker, R.B. Silberstein, P.J. Cadusch, EEG coherency I: statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales, Electroenceph. Clin. Neurophysiol. 103 (1997) 499–515. [20] M. Essl, P. Rappelsberger, EEG coherence and reference signals: experimental results and mathematical explanations, Med. Biol. Eng. Comput. 36 (1998) 1–8. [21] L. Leocani, C. Toro, P. Manganotti, P. Zhuang, M. Hallett, Event-related coherence and event related desynchronization/synchronization in the 10 Hz and 20 Hz EEG during self-paced movements, Electroenceph. Clin. Neurophysiol. 104 (1997) 199–206. [22] J.M. Ford, D.H. Mathalon, S. Whitfield, W.O. Faustman, W.T. Roth, Reduced communication between frontal and temporal lobes during talking in schizophrenia, Biol. Psychiatry. 51 (6) (2002) 485–492. [23] T. Locatelli, M. Cursi, M. Franceschi, G. Comi, EEG coherence in Alzheimer’s disease, Electroenceph. Clin. Neurophysiol. 106 (1998) 229–237. [24] J. Gallinat, G. Winterer, C.S. Herrmann, D. Senkowski, Reduced oscillatory gamma-band responses in unmedicated schizophrenic patients indicate impaired frontal network processing, Clin. Neurophysiol. 115 (8) (2004) 1863–1874. Xiaofeng Liu was born in Shanxi, PR China, in 1974. He received his B.S. degree in Electronic Engineering and M.S. degree in Computer Application from Taiyuan University of Technology, China, in 1996 and 1999, respectively. At present he is a Ph.D. student in the Department of Biomedical Engineering of Xi’an Jiaotong University. His current research interests include biomedical signal processing and neuroimaging analyses of language processing and selective attention. Huan Qi was born in Shaanxi, PR China, in 1981. He received his B.S. degree from the Department of Biomedical Engineering of Xi’an Jiaotong University. Currently he is a postgraduate student to attain the M.S. degree in the Department of Biomedical
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1315
Engineering of Xi’an Jiaotong University. His current research interests are the time-frequency/time-scale analyses of evenrelated brain potentials. Supin Wang was born in Hebei, PR China, in 1950. She received a diploma in Electrical Engineering from Xi’an Jiaotong University of China in 1976. She became a full professor in the Department of Biomedical Engineering, Xi’an Jiaotong University in 1999. She has authored and co-authored more than 30 publications and three books. Her research interests include biomedical signal processing and biomedical ultrasound. Mingxi Wan was born in Hubei, PR China, in 1962. He received his B.S. degree in Geophysical Prospecting in 1982 from Jianghan Petroleum Institute and his M.S. and Ph.D. degrees in Biomedical Engineering from Xi’an Jiaotong University, China, in 1985 and 1989, respectively. He is now a Professor and Chairman of the Department of Biomedical Engineering at Xi’an Jiaotong University. He has authored and co-authored more than 80 publications and three books on biomedical ultrasound. He has received several important awards from the Chinese government and university. His research interests include voice science and biomedical ultrasound.
Wavelet-based estimation of EEG coherence during Chinese Stroop task Xiaofeng Liu, Huan Qi, Supin Wang, Mingxi Wan∗ The Key Laboratory of Biomedical Information Engineering of Ministry of Education, and Department of Biomedical Engineering, School of Life Science and Technology, Xi’an Jiaotong University, Xi’an 710049, PR China Received 16 May 2005; accepted 11 August 2005
Abstract Wavelet-based estimation of instantaneous EEG coherence was applied to investigate the synchronization of different brain regions whilst 10 subjects performing Stroop task presented in Chinese. In contrast to coherence based on Fourier transform, wavelet-based coherence, which does not depend on the stationarity of signals, applies an adaptive window to the frequency of the signal and has a more accurate time-frequency resolution. In the present study, a greater negativity for the incongruent situation than congruent situation appeared from 350 to 600 ms post-stimulus onset over frontal, central, and parietal regions, and significantly higher EEG coherences for the incongruent situation than congruent situation were observed over frontal, parietal, and frontoparietal regions from 100 to 400 ms at 1 (13–18 Hz) frequency band, which was found to be sensitive in the discrimination between congruent and incongruent situations. The findings in the present study may indicate that functional synchronization as indexed by EEG coherence at 1 frequency band is enhanced at the earlier stage while processing the conflicting information from the incongruent stimulus, and that 1 frequency band is close related to interactions of brain areas in the selected attention task. 䉷 2005 Elsevier Ltd. All rights reserved. Keywords: EEG; Coherence; Wavelet; Chinese Stroop task; Functional synchronization
∗ Corresponding author. Tel.: +86 29 82667924; fax: +86 29 82668668.
E-mail address: [email protected] (M. Wan). 0010-4825/$ - see front matter 䉷 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiomed.2005.08.002
1304
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1. Introduction The coherence of scalp recorded EEG is a frequency domain measure of similarity between EEG signals and gives a measure of the degree of functional cooperation between neuronal systems underlying the generation of those signals [1]. The EEG coherence has proven itself a promising tool for investigating the synchronization of the cortical activity during various cognitive processes [2–6]. The Stroop task [7], in which subjects are required to name the color of color-word printed in congruent (e.g., RED in red) or incongruent (e.g., RED in blue) ink as quickly as possible while ignoring the meaning of word, involves selective attention, language processing, color-naming processes. Therefore, the investigation of the synchronization of functional networks during Stroop task by means of coherence may be helpful to understand the neurophysiological basis of high-level cognitive activities of normal humans, and even to provide clinical assessments for dysfunctions or disorders in patients. In numerous studies on the Stroop task, event-related potential (ERP) is often employed to explore the temporal course of the related neural processes because of its capability to record brain activity with a millisecond level temporal resolution. Recently, the electrophysiological correlates of the Stroop task with three different response modalities (Overt Verbal, Covert Verbal, and Manual) were investigated [8]. It was found that a negativity, the negative-going ERP waveform, was greater for the incongruent situation than congruent situation at 350–500 ms post-stimulus onset (peaking at 410 ms) over midline sites, which were interpreted as evidences of conflict processing and resolution in the anterior cingulate. In addition, a greater positive shift of ERP for the incongruent situation than congruent situation was observed at 500–800 ms over the left temporo-parietal regions and negativity over the anterior frontal region. Another ERP study [9] of Stroop effect also reported a greater negativity for the incongruent situation than congruent situation over the fronto-central region beginning at about 500 ms and a greater positive shift over the left temporo-parietal regions beginning at approximately 650 ms. In the previous studies, the neural activity associated with Stroop task over the independent electrode site was mainly researched. Consequently, the temporal course and distribution of the neural activity were analyzed and reported in detail, but not the functional interaction of the different areas. Nowadays, it is widely held that different regions of the brain must cooperate with each other to support integration of sensory information, sensory-motor coordination and many other functions that are critical for learning, memory, information processing, perception and the behavior of organisms [10]. West and Bell [11] observed an increased alpha-1 (8–10 Hz) within the prefrontal and parietal areas when they examined EEG power for different frequency bands over the frontal and parietal regions in a Stroop study, and assumed an interaction between prefrontal and parietal regions. Schack et al. [12] firstly addressed the interaction processes of different brain regions in the Stroop task in relation to the time-evolution of the cognitive process by means of instantaneous coherence analysis based on Fourier transform (FT). Their research revealed that the frequency band of 13–20 Hz was sensitive to the discrimination between the congruent and incongruent situations, and that higher coherences were observed within the left frontal and left parietal areas, as well as between them for the incongruent situation in comparison with the congruent situation. Unfortunately, vocal response modality in Stroop task was employed, and thus, the facial muscle movements underlying speech production could introduce a significant artifact into EEG. In the present study, the neural correlates of Chinese Stroop task are addressed by using high-density EEG recordings. Then, the properties of coherence patterns of different brain areas in different conditions
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1305
during Stroop task are examined by means of wavelet-based coherence that has been often applied to the estimation of coherence among non-stationary signals [13–16]. Wavelet-based coherence has a more accurate time-frequency resolution than the coherence based on the windowed short-term Fourier transform, in that the size of the window is adapted to the frequency of the signal in wavelet analysis, while it is fixed in short-term Fourier transform; thus, it provides the time-varying power spectrum as well as the phase spectrum.
2. Materials and methods 2.1. Subjects Ten healthy native Chinese (aged 21–30, mean = 23.4 years, five males and five females) participated in this study with informed consent. All subjects were right handed according to a self-report, and free from neurological or psychiatric illness. All had normal or corrected to normal visual acuity, and normal color vision. 2.2. Stimuli and procedure Stimuli (Chinese equivalents of Red, Green and Blue) were printed in the color red, or green, or blue and presented in the center of the screen against a black background. The stimulus subtended approximately 4.9◦ visual angle horizontally and 1.4◦ vertically. Subjects were seated inside a dimly lit room at a distance of approximately 70 cm from the screen. The experiment included a color-to-key acquisition phase, a practice phase, and a test phase. In the color-to-key acquisition, subjects performed a block of 60 trials (XXX was printed in each of the three colors appeared 20 times) to establish a strong mapping between the stimulus color and corresponding key. Also, a practice task of 36 trials (color-word stimulus) was completed to acquaint the subject with the task. Each subject participated two blocks of 90 trials (equivalent trials for congruent and incongruent). Each trial of a block was pseudo-randomized presented. Subjects were instructed to pay attention to the word presented on the screen and response to ink color of the words as quickly and accurately as possible by pressing the appropriate key. Stimuli duration was 100 ms, and stimulus onset asynchrony (SOA) varied randomly between 1600 and 1800 ms to minimize the contribution of expectation effects on the stimulus-locked EEG. The block presentation order was counterbalanced across subjects. 2.3. EEG recording EEG activity was continuously recorded from the scalp using EEG amplifier (SynAmps, NeuroScan Inc., USA) with a 32-channel electrode-cap (Quik-Cap). All electrodes were referenced to linked electrode placed on the left and right mastoids (International 10/20-Systems). EEG channels were amplified using a band-pass of 0.1–45 Hz, sampling frequency at 250 Hz, impedence below 5 K, and a gain of 12 500. Electrooculogram (EOG) activity was recorded from four electrodes (two placed above and below the left eye; two placed at external canthus of each eye) and used as a guide to removal of eye movement artifacts and blinks. ERP analysis epochs were extracted offline and included 100 ms pre-stimulus and
1306
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
900 ms post-stimulus. Any trial on which EOG or EEG amplitude exceeded ±100 V was rejected as artifact. Only EEGs of trials associated with a correct response and not contaminated with artifacts were averaged according to stimulus type. 2.4. Wavelet coherence The continuous wavelet transform of a square integrable complex-valued signal x(t) is defined as Wx (, f ) =
+∞ −∞
x(t)∗,f (t) dt.
(1)
Morlet wavelet is simple and well suited for spectral estimations. It is defined for frequency f and time by (u − )2 , (2) ,f (u) = f exp i2f (u − ) exp − 2
,f (u) is simply the product of a sinusoidal wave at frequency f, with a Gaussian function centered at time with a standard deviation proportional to the inverse of f. From the wavelet transforms of two signals x(t) and y(t), the wavelet auto-spectrum of x(t) and/or y(t) and wavelet cross-spectrum between x(t) and y(t) are defined as w Sxx (t, f ) = Wx (, f )Wx∗ (, f ) d, (3) w Sxy (t, f ) =
T
T
Wx (, f )Wy∗ (, f ) d.
(4)
The localized time integration window in Eqs. (3 and 4) T = [t − , t + ] is selected based on the time resolution desired in the resulting coherence map ( can be adapted to the frequency of interest); thus, the coherence map provides a view of localized correlation with respect to both time and frequency. Finally, the wavelet coherence w xy (t, f ) is defined at time t and frequency f by w xy (t, f ) =
w (t, f )| |Sxy
(5)
w (t, f )S w (t, f ) Sxx yy
w (t, f ), a quantity bounded by zero and unity, The magnitude squared coherence (MSC), denoted by Cxy is defined at time t and frequency f by w 2 Cxy (t, f ) = |w xy (t, f )| = w 0 Cxy (t, f ) 1 ∀f .
w (t, f )|2 |Sxy
w (t, f )S w (t, f ) Sxx yy
.
(6a) (6b)
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1307
The average of the wavelet coherence w xy (t, f ) between frequency bands enables a necessary data reduction. As a result, mean band coherence as a function of time is obtained. For a chosen frequency band [fl , fu ] the mean band coherence is computed by fu w (t) = w (7a) xy (t, f ) df , f fl
2 Cfw = |w f| .
(7b)
Further, mean coherence within a chosen time-frequency region [t1 , t2 ] ∗ [fl , fu ] is defined at time t by t2 f u w (t) = w (8a) xy (t, f ) df dt, t,f t1
fl
w 2 = |w Ct,f t,f | .
(8b)
The coherence analysis was performed for adjacent electrode pairs within the frontal, central, temporal, and parietal regions and additionally for frontoparietal electrode pairs, such as F3/P3, F3/Pz, F3/P4, Fz/P3, Fz/Pz, Fz/P4, F4/P3, F4/Pz, F4/P4, and temporoparietal electrode pairs, such as TP7/P3, P3/T7, TP7/T7, TP8/P4, P4/T8, and TP8/T8. 3. Results 3.1. Event-related potentials Fig. 1 showed the grand average ERP elicited by congruent and incongruent trials for correct responses. All trials elicited a late positive complex wave with maximum over the midline centro-parietal region. A greater negativity for the incongruent situation than congruent situation, peaking at around 440 ms, was observed between 350 and 600 ms post-stimulus onset over frontal, central, and parietal regions (see Fig. 1). The current results were consistent with previous ERP studies [8,9]. 3.2. Coherence analysis 3.2.1. Time frequency map of EEG coherence in congruent and incongruent task situations Due to wavelet transform, it is possible to examine the instantaneous EEG coherences at different frequencies by means of time-frequency representations (TFRs). In the present study, wavelet coherence was performed for the electrode pairs Fp1/F3, Fp1/F7, F7/F3, F3/Fz and F7/Fz within the left frontal area, for the electrode pairs Fp2/F4, Fp2/F8, F8/F4, F4/Fz and F8/Fz within the right frontal area, for the electrode pairs TP7/P3, P3/T7, and TP7/T7 within the left temporoparietal area, for the electrode pairs TP8/P4, P4/T8, and TP8/T8 within the right temporoparietal area and for the frontoparietal electrode pairs F3/P3, F3/Pz, F4/P4 and F4/Pz. We estimated wavelet coherence separately for all the single trials, and then averaged the resulting distributions to enhance time-frequency regions bearing similar timefrequency signatures. As a result, higher coherences were found over the electrode pairs within and between the frontal and the parietal regions. Fig. 2 showed the time-frequency map of wavelet coherence of the electrode pair F3/Pz in the congruent and incongruent situations and the difference between them
1308
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 F3
F1
0
0
4
4 0
400 ms
800
0
C3
400 ms
800
C1
0
0
4
4 0
400 ms
0
800
CZ
400 ms
800
PZ
0
0
4
4 0
400 ms
800
0
T7 0
4
4 400 ms
800
T8
0
0
400 ms
800
0
400 ms
800
Fig. 1. Grand average ERPs over 10 subjects at selected electrode sites F3, F4, C3, C4, Cz, Pz, T7, and T8 for the congruent (bold real line) and incongruent (thin dotted line) conditions.
(incongruent minus congruent situation), averaged for all single trials and all 10 subjects. To make it clear and enhance the comparison, we selected out the values higher than the threshold (herein, 0.5) of the differences between the incongruent and congruent situations, which were shown in the lower right panel of Fig. 2. In Fig. 2, higher coherences were observed at frequency band of beta (13–30 Hz) and gamma (> 30 Hz) for both incongruent and congruent situations, and higher difference between the incongruent and congruent situations appeared at the beta frequency band before around 400 ms post-stimulus onset. In order to objectively assess the degree of coherence between different electrode pairs, the mean wavelet coherence was calculated for all 10 subjects and both task conditions. Generally, high coherences appeared during the whole task for higher frequency bands, beta (13–30 Hz) and gamma (> 30 Hz) frequency bands, for the two task situations, and mainly distributed within frontal, parietal, frontal-central, and fronto-parietal regions. According to the TFRs of the coherence difference (incongruent minus congruent situation) within and between the frontal, central, parietal, and temporal regions, salient coherence difference appeared in the time-frequency region, approximately at [100–400 ms]∗[10–20 Hz], for electrode pairs within and between frontal and parietal regions, such as Fp1/F3, Fp1/F7, F7/F3, F3/Fz, Fp2/F4, Fp2/F8, F8/F4, F4/Fz, P3/Pz, P4/Pz, F3/P3, F3/Pz, F4/P4, and F4/Pz, namely, the highest differences between congruent and incongruent situations appeared approximately within this time-frequency region. To further assess the sensitivity of different frequency bands for the discrimination of the two task situations, the mean band coherences for standard frequency bands were calculated according to Eq. (7b). The mean time courses of band coherences for the electrode pair F3/Pz were shown in Fig. 3. The largest
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 Congruent
1309
InCongruent 1.0
1
10
10
0.75
0.75
20 Hz
Hz
20 0.50
30 0.25
40 50
0.50 30
0.0 0
200
400 ms
600
0.25
40 50
800
0
Incongruent-Congruent Difference
200
400 ms
600
800
0.0
Selected Difference (> 0.5) 1.0 0.6
10
10 0.75
0.3 20 0.0 30
Hz
Hz
20 0.50 30 -0.3
40
0.25
40 -0.6
50
0
200
400 ms
600
800
50
0
200
400 ms
600
800
0.0
Fig. 2. Time-frequency map of wavelet coherence of the electrode pair F3/Pz averaged for all single trials and all 10 subjects. Upper left panel: Congruent; Upper right panel: Incongruent; Lower left panel: Incongruent minus Congruent difference; Lower right panel: Selected difference between incongruent and congruent situations (only values higher than 0.5 were kept). Color-scale from ‘black’ (minimum) to ‘white’ (maximum).
coherence difference between incongruent and congruent situations appeared at the 1 frequency band (13–18 Hz) for the electrode pair F3/Pz in Fig. 3, so did they for frontal and parietal areas in the left hemisphere. The mean time courses of band coherences averaged over the electrode pairs F7/F3, F7/Fz, F3/Fz, F3/P3, F3/Pz, F7/P3, F7/Pz, and P3/Pz in the left hemisphere were represented in Fig. 4. The coherence difference between incongruent and congruent situations at the 1 frequency band for left hemispherical frontal and parietal areas lasted for longer time than that for the single electrode pair F3/Pz. A similar effect was also found for the coherences averaged over the electrode pairs in the right hemisphere. The mean band coherences averaged over the electrode pairs F8/F4, F8/Fz, F4/Fz, F4/P4, F4/Pz, F8/P4, F8/Pz, and P4/Pz of the right hemisphere were shown in Fig. 5 . On this basis, the frequency band of 13–18 Hz was roughly deemed sensitive to discriminate the congruent and incongruent task situations, and highest differences of coherence between the two situations mainly occurred at around 100–400 ms.
1310
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.6
0.2
0.4
0
0.2
0
200
400
600
-0.2
0
200
ms
(a)
400
600
400
600
400
600
400
600
ms 0.2
0.8
0.1 0.6 0 0.4
0
200
(b)
400
600
-0.1
0
200
ms
ms
1
0.2 0.1
0.8 0 0.6
-0.1 0
200
(c)
400
600
0
200
ms
ms
1 0.1 0.8
0 -0.1
0.6 (d)
0
200
400 ms
600
0
200 ms
Fig. 3. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent situations (gray real line, right panel) for the electrode pair F3/Pz for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz. The thin vertical line denotes the mean reaction time for the congruent situation (483 ms), the thick vertical line denotes the mean reaction time for the incongruent situation (527 ms).
3.2.2. Topographical map of EEG coherence at a sensitive time-frequency region As described in the preceding section, the frequency band of 13–18 Hz had proven to be sensitive for discrimination between the congruent and incongruent task situations within 100–400 ms. The mean coherences within the time-frequency region, [13–18 Hz]∗[100–400 ms], over these electrode pairs within and between frontal, central, temporal, parietal, and occipital regions, were calculated according to Eq. (8b), and then the topography of coherences of the two task situations was obtained and illustrated in Fig. 6. The mean coherence averaged over Fp1/F7, Fp1/F3, F3/F7, F3/Fz, F7/Fz, F3/P3, F3/Pz, F7/P3, and F7/Pz represents the left hemisphere and the mean coherence averaged over Fp2/F8, Fp2/F4, F4/F8, F4/Fz, F8/Fz, F4/P4, F4/Pz, F8/P4, and F8/Pz represents the right hemisphere, analogously. There was no significant difference in coherence between the hemispheres. In addition, coherences for the incongruent situation were significantly stronger than those for the congruent situation in not only the left hemisphere
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.8
0.2
0.6
0
0.4 (a)
0
200
400
600
-0.2
0
200
ms
1311
400
600
400
600
400
600
400
600
ms 0.2
0.8
0.1 0
0.6 (b)
0
200
400
600
-0.1
0
200 ms
ms 0.2
0.8
0.1 0
0.6 (c)
0
200
400
600
-0.1
0
200
ms
ms 0.2
0.8
0.1 0
0.6 0 (d)
200
400 ms
600
-0.1
0
200 ms
Fig. 4. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent situations (gray real line, right panel) for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz averaged over the left hemispherical electrode pairs F7/F3, F7/Fz, F3/Fz, F3/P3, F3/Pz, F7/P3, F7/Pz, P3/Pz, F3/O1, F3/Oz, F7/O1, F7/Oz.
but also the right hemisphere, as shown in Fig. 6. Therefore, the synchronization related to the Stroop effect seems to be symmetrical according to the coherence of the sensitive frequency band 13–18 Hz.
4. Discussion The aforementioned coherence results were achieved with linked mastoid reference. Although, the reference electrode’s influence on EEG coherence was demonstrated in several studies [17–20], the comparison between the congruent and incongruent task situations could be performed on the basis of instantaneous coherence, as Schack and his colleagues argued [12]. EEG coherence analysis, an important tool for studying high-level cognitive processes, enables us to examine the intermediate sub-processes of interaction among different topographic areas and thus gives insights into the investigation of the functional networks cooperation during various cognitive processes.
1312
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 0.8
0.2
0.6
0
0.4 (a)
0
200
400
600
-0.2
0
200
ms
400
600
400
600
400
600
400
600
ms
0.8 0.2 0.6
0.1
0.4
-0.1
0 0
200
(b)
400
600
0
200 ms
ms
0.8
0.2 0.1 0
0.6 (c)
0
200
400
600
-0.1
0
200
ms
ms 0.2
0.8
0.1 0
0.6
-0.1 0
(d)
200
400 ms
600
0
200 ms
Fig. 5. Mean time courses of frequency band coherences of the incongruent (bold real line, left panel), the congruent (thin dashed line, left panel) situations, and the coherence difference between incongruent and congruent (gray real line, right panel) for the frequency bands: (a) 8–12, (b) 13–18, (c) 18–30, (d) 30–40 Hz averaged over the right hemispherical electrode pairs F8/F4, F8/Fz, F4/Fz, F4/P4, F4/Pz, F8/P4, F8/Pz, P4/Pz, F4/O2, F4/Oz, F8/O2, F8/Oz.
Therefore, coherence between bioelectrical signals has been frequently applied to measure the relationship between EEG/MEG signals recorded during cognitive function of normal human [2–5,21] and psychiatric or neurological patients [22–24]. The wavelet coherence is very different from the traditional spectral coherence, in that it describes the correlation of multiscale wavelet components rather than fixed-scale sinusoidal components of the EEG signals. In the present study the wavelet coherence was applied to address the interference during incongruent Stroop task. High coherences appeared during the whole task for frequency band beta (13–30 Hz) and gamma (> 30 Hz) for the two situations in the present study. The coherences within time interval 100–400 ms at the frequency band 13–18 Hz were overall higher for incongruent situations than those for congruent situations over the electrode pairs within the frontal, central, and parietal regions, and additionally for frontoparietal regions. The frequency band beta-1 (13–18 Hz) was found sensitive to the discrimination between the incongruent and congruent situation, which was consistent with the research of Schack and his colleagues [12]. Coherences for the incongruent situation were significantly stronger than those for
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315 Congruent
Incongruent
Fp2
Fp1
T7
TP7
FP1 F8
F7
FT7
F3
Fz
F4
FC3
FCz
FC4
C3
Cz
C4
CP3
CPz
CP4
P3
Pz Oz
O2
F8 Fz
F4
FCz
FC4
C2
Cz
C4
CP3
CPz
CP4
P3
Pz
P4
O1
Oz
O2
F3 FT8
FP2
P7
PT7
T8
T7
TP8
TP7
P9
P7
P4
P7 O1
1313
C3
FT8
T8
TP8
P9
Fig. 6. Topographical maps of the mean coherence at frequency band 13–18 and time interval 100–400 ms (weight of line is proportional to magnitude of coherence).
the congruent situation in both the left hemisphere and the right hemisphere. Therefore, it seemed to be plausible that interaction or synchronization as indexed by EEG coherence between the functional brain cortical areas was enhanced when processing the incongruent stimulus, according to the higher coherence for the incongruent situation than congruent situation. In the present study, the Stroop effect seems to be symmetrical according to the synchronization as indexed by coherence, which is distinct to the findings of Schack et al. [12] who observed a left lateralized synchronization phenomenon. This may be attributed to the difference in the processing of the Chinese stimulus and English stimulus, and needs further investigation. Acknowledgements The authors gratefully acknowledge the support from the National Natural Science Foundation of China (Grants 30070212 and 69925101). References [1] P.L. Nunez, EEG coherence measures in medical and cognitive science: a general overview of experimental methods, computer algorithms, and accuracy, in: H. Witte, U. Zwiener, B. Schack, A. Doering (Eds.), Quantitative and Topological EEG and MEG Analysis, Druckhaus Mayer, Jena, 1997, pp. 385–392. [2] H. Petsche, S.C. Etlinger, O. Filz, Brain electrical mechanism of bilingual speech management: an initial investigation, Electroenceph. Clin. Neurophysiol. 86 (1993) 385–394. [3] H. Petsche, Approaches to verbal, visual and musical creativity by EEG coherence analysis, Int. J. Psychophysiol. 24 (1996) 145–159. [4] H. Petsche, S.C. Etlinger, EEG and Thinking, Oesterreichische Akademie der Wissenschaften, Wien, 1998. [5] J. Sarnthein, H. Petsche, P. Rappelsberger, G.L. Shaw, A. von Stein, Synchronization between prefrontal and posterior association cortex during human working memory, Proc. Natl. Acad. Sci. USA 95 (1998) 7092–7096.
1314
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
[6] S. Weiss, H. Mueller, The contribution of EEG coherence to the investigation of language, Brain Lang. 85 (2003) 325–343. [7] J.R. Stroop, Studies of interference on serial verbal reactions, J. Exp. Psychol. 18 (1935) 643–662. [8] M. Liotti, M.G. Woldorff, R. Perez, H.S. Mayberg, An ERP study of the temporal course of the Stroop color-word interference effect, Neuropsychologia 38 (2000) 701–711. [9] R. West, C. Alain, Event-related neural activity associated with the Stroop task, Cognitive Brain Res. 8 (1999) 157–164. [10] W.H.R. Miltner, C. Braun, M. Arnold, H. Witte, E. Taub, Coherence of gamma-band EEG activity as a basis for associative learning, Nature 397 (1999)434–436. [11] R. West, M.A. Bell, Stroop color-word interference and electroencephalogram activation: evidence for age-related decline of the anterior attention system, Neuropsychology 11 (1997) 421–427. [12] B. Schack, A.C. Chen, S. Mescha, H. Witte, Instantaneous EEG coherence analysis during the Stroop task, Clin. Neurophysiol. 110 (1999) 1410–1426. [13] P. Liu, Wavelet spectrum analysis and ocean wind waves, in: E. Foufoula-Georgiou, P. Kumar (Eds.), Wavelets in Geophysics, Academic Press, New York, 1994, pp. 151–166. [14] S. Santoso, E. Powers, R. Bengtson, A. Ouroua, Time-series analysis of nonstationary plasma fluctuations using wavelet transforms, Rev. Sci. Instrum. 68 (1997) 898–901. [15] K. Gurley, T. Kijewski, A. Kareem, First- and higher-order correlation detection using wavelet transforms, J. Eng. Mech. 129 (2) (2003) 188–201. [16] J.P. Lachaux, A. Lutz, D. Rudrauf, D. Cosmelli, M. Le Van Quyen, J. Martinerie, F. Varela, Estimating the time-course of coherence between single-trial brain signals: an introduction to wavelet coherence, Clin. Neurophysiol. 32 (2002) 157–174. [17] C. Andrew, G. Pfurtscheller, Dependence of coherence measurements on EEG derivation type, Med. Biol. Eng. Comput. 34 (1996) 232–238. [18] T.D. Lagerlund, F.W. Sharbrough, N.E. Busacker, K.M. Cicora, Interelectrode coherences from nearest-neighbor and spherical harmonic expansion computation of Laplacian of scalp potential, Electroenceph. Clin. Neurophysiol. 95 (1995) 178–188. [19] P.L. Nunez, R. Srinivasan, R.S. Wijesinghe, A.F. Westdorp, D.M. Tucker, R.B. Silberstein, P.J. Cadusch, EEG coherency I: statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales, Electroenceph. Clin. Neurophysiol. 103 (1997) 499–515. [20] M. Essl, P. Rappelsberger, EEG coherence and reference signals: experimental results and mathematical explanations, Med. Biol. Eng. Comput. 36 (1998) 1–8. [21] L. Leocani, C. Toro, P. Manganotti, P. Zhuang, M. Hallett, Event-related coherence and event related desynchronization/synchronization in the 10 Hz and 20 Hz EEG during self-paced movements, Electroenceph. Clin. Neurophysiol. 104 (1997) 199–206. [22] J.M. Ford, D.H. Mathalon, S. Whitfield, W.O. Faustman, W.T. Roth, Reduced communication between frontal and temporal lobes during talking in schizophrenia, Biol. Psychiatry. 51 (6) (2002) 485–492. [23] T. Locatelli, M. Cursi, M. Franceschi, G. Comi, EEG coherence in Alzheimer’s disease, Electroenceph. Clin. Neurophysiol. 106 (1998) 229–237. [24] J. Gallinat, G. Winterer, C.S. Herrmann, D. Senkowski, Reduced oscillatory gamma-band responses in unmedicated schizophrenic patients indicate impaired frontal network processing, Clin. Neurophysiol. 115 (8) (2004) 1863–1874. Xiaofeng Liu was born in Shanxi, PR China, in 1974. He received his B.S. degree in Electronic Engineering and M.S. degree in Computer Application from Taiyuan University of Technology, China, in 1996 and 1999, respectively. At present he is a Ph.D. student in the Department of Biomedical Engineering of Xi’an Jiaotong University. His current research interests include biomedical signal processing and neuroimaging analyses of language processing and selective attention. Huan Qi was born in Shaanxi, PR China, in 1981. He received his B.S. degree from the Department of Biomedical Engineering of Xi’an Jiaotong University. Currently he is a postgraduate student to attain the M.S. degree in the Department of Biomedical
X. Liu et al. / Computers in Biology and Medicine 36 (2006) 1303 – 1315
1315
Engineering of Xi’an Jiaotong University. His current research interests are the time-frequency/time-scale analyses of evenrelated brain potentials. Supin Wang was born in Hebei, PR China, in 1950. She received a diploma in Electrical Engineering from Xi’an Jiaotong University of China in 1976. She became a full professor in the Department of Biomedical Engineering, Xi’an Jiaotong University in 1999. She has authored and co-authored more than 30 publications and three books. Her research interests include biomedical signal processing and biomedical ultrasound. Mingxi Wan was born in Hubei, PR China, in 1962. He received his B.S. degree in Geophysical Prospecting in 1982 from Jianghan Petroleum Institute and his M.S. and Ph.D. degrees in Biomedical Engineering from Xi’an Jiaotong University, China, in 1985 and 1989, respectively. He is now a Professor and Chairman of the Department of Biomedical Engineering at Xi’an Jiaotong University. He has authored and co-authored more than 80 publications and three books on biomedical ultrasound. He has received several important awards from the Chinese government and university. His research interests include voice science and biomedical ultrasound.