禪坐腦電波之時變頻譜的空間特性研究(I)

行政院國家科學委員會補助專題研究計畫成果報告

禪坐腦電波之時變頻譜的空間特性研究(I)

Time-varying Spatio-spectral Characteristics of Meditation EEG

計畫類別：■個別型計畫 □整合型計畫

計畫編號：NSC 96-2221-E-009-217

執行期間：96 年 08 月 01 日 至 97 年 07 月 31 日

計畫主持人：羅佩禎

共同主持人：

成果報告類型（依經費核定清單規定繳交）：█精簡報告 □完整報告

本成果報告包括以下應繳交之附件：

□赴國外出差或研習心得報告一份

□出席國際學術會議心得報告及發表之論文各一份

□國際合作研究計畫國外研究報告書一份

處理方式： 除產學合作研究計畫、提昇產業技術及人才培育研究計畫、列

管計畫及下列情形者外，得立即公開查詢

□涉及專利或其他智慧財產權，█一年□二年後可公開查詢

執行單位：國立交通大學 電機與控制工程學系

中 華 民 國 九十七 年 十 月 二十七 日

行政院國家科學委員會專題研究計畫成果報告

禪坐腦電波之時變頻譜的空間特性研究(I)

Time-varying Spatio-spectral Characteristics of Meditation EEG

計畫編號： NSC 96-2221-E-009-217

執行期限： 96年08月01日至97年7月31日

主 持 人： 羅佩禎，國立交通大學電機與控制工程學系

中文摘要

過去十年來，主持人將醫學工程研究經驗投入於禪坐之生理、意識等現象的探討， 已有相當成果。主要以科學化方法來探討禪定過程中之腦電波特性變化，以禪宗佛法之 修行者為主要研究對象；以先進之數位訊號處理的方法理論，從大量記錄、收集之多通 道禪定腦電波中，進行時間、頻率分析。其中觀察到某些特性，值得進一步探究其腦部 動態機制與空間定位之關聯性。 為瞭解此種非睡眠、超意識狀態下之腦電波活動情形；並了解禪坐中其他生理指標 的關聯性，本研究計畫分兩年探討：（一）禪坐腦電波之空間特性定量(MBM)，（二） 禪坐過程中之MBM 演變情形(MBMS)，（三）MBM 與其他生理現象（如心律變異、 皮膚阻抗）的相關性。研究過程中，除了進行大量腦電波記錄實驗，亦將發展多元化之 數位訊號分析方法，來萃取和量化波形特徵，並能有系統的建立一受測者資料庫。在過 去 這 一 年 的 計 畫 中 ， 我 們 巳 經 使 用 頻 譜 分 析 和 同 調 性(coherence) 分 析 的 腦 殼 圖 (topographic map)，完成初步的禪坐腦電波之時空特性研究。 關鍵詞：禪坐之生理與意識現象、多通道腦電波、時變頻譜之空間特性、腦電波之同調 性。 ABSTRACT

For more than ten years, the principal investigator has been devoted to the research on physiological and mental/conscious phenomena under Zen meditation. A number of important results have been reported, of which we mainly focus on investigating the EEG (electroencephalograph) characteristics based on the scientific approach. Subjects of the experiment practice the Zen Buddhism. From a large amount of meditation EEG signals acquired, we characterized their temporal and spectral features by a number of advanced DSP

foci that generate such kind of Zen brain dynamics.

To understand the phenomena of brain electrical activities and other relevant physiological parameters under such a non-sleeping, transcendental state, this two-year research proposal is aimed at: (1) quantitative analysis of meditation brain mapping (MBM), (2) time-varying MBMs, or the MBM Scenario (MBMS), and (3) correlation between MBM and other physiological parameters (for example, heart rate variability HRV and galvanometric skin resistance GSR). In the first-year research, we have applied the topographic maps of relative power and coherence to investigate the spatiotemporal characteristics of Meditation EEG.

Keywords: Physiological state and consciousness under Zen meditation, multi-channel EEG

(electroencephalograph), Time-varying spatio-spectral EEG characteristics, EEG coherence.

INTRODUCTION

Since the interaction between separate brain regions is neurophysiologically very significant and appealing, the study on synchronous oscillations also plays an important role in the study of meditation EEG. Coherence function, a generalization of correlation in the frequency domain, is the most popular tool for the assessment of ‘interaction’ at a specific frequency (Nunez et al., 1999). EEG coherence is a measure of the synchronization between the two recording sites and may be interpreted as an expression of their functional interaction. Spatiotemporal characteristics of Meditation EEG were investigated using EEG coherence.

In Zen-meditation practice, one major technique of attaining good-quality meditation is to guard and focus on the particular site (called Chakra), for example, Zen Chakra locating inside the third ventricle (refer to Introduction of Zen Meditation, lectured by Master Wu Chueh Miao Tien, published by Zen Cosmos, 2004). By means of Chakra focusing,

experienced practitioners reported that they often could enter into a tranquil state of consciousness transcending beyond the physiological (the fifth), mental (the sixth), subconscious (the seventh), and Alaya (the eighth) conscious state (Chang & Lo, 2005). This study investigates EEG coherence in the phases of meditation and Zen-Chakra guarding as comparison with rest. The aims of the work are to obtain if there are changes of EEG coherence between left versus right hemispheres and frontal versus occipital cortex during meditation, to evaluate whether short-distance or long-distance coherence is more affected.

METHOD AND MATERIAL

1. Subjects

In this study, twelve Zen-meditation practitioners (mean age: 32.5 year; range: 28–41 years; four females; all right-handed) were investigated. Since reliable effects of meditation on the EEG activity were obtained mainly in experienced meditators, all were long-term practitioners with an average experience of 8.8 years (range 5 to 13 years) in Zen-Buddhist meditation practice.

2. Procedure

Previous studies on hundreds of practitioners revealed large between-group variations in EEG pattern. We thus aimed to investigate the within-subject behaviors during the entire meditation course. In the beginning 3 minutes, we collected the baseline data for each meditator under normal, eye-closed rest (R), without entering into meditation. Then they were instructed to begin a 40-min meditation. After the meditation, the subjects were asked to keep their eyes closed for an additional 3-min transition period. Then they were asked to practice a 5-min Zen-Chakra focusing. During the recording period, the subjects sat in the full-lotus position with eyes closed. After the recording, the subjects were asked for an interview. M session indicates a 3-min artifact-free epoch extracted from the beginning after 30-min of

meditation. Z session refers to the first 3-min artifact-free data segments under Zen-Chakra focusing.

3. Recording details

The EEG signals were recorded by the 30-channel, common-reference (linked-mastoid MS1-MS2) electrode montage based on the international 10-20 system. As the number of coherence measure increases with the square of channel number, a data reduction was required. For simplification, EEG coherence was computed for 19 scalp locations from 30 channel EEG recordings as showed in Fig. 1. EEG signals were sampled at 200 Hz after 0.3-45 Hz band-pass filtering. Each EEG epoch (R, M and Z) was analyzed by running measurement of nonlinear interdependence using a 5-second window without overlap.

4. EEG coherence analysis

Coherence between two signals x and y is the ratio between their cross-spectral density and their individual auto-spectral density. However, due to finite size of the neural data, one can only have an estimate of the true spectrum. Smoothing techniques are often used to improve the performance of the spectral estimators. Thus, we estimate the coherence function using Welch's averaged, modified periodogram method (Welch, 1967). EEG signals are usually subdivided into M equal overlapping sections, so that coherence is calculated as:

) ( ) ( ) ( 2 2 f S f S f S C yy xx xy xy = (1)

where〈·〉indicates average over the M segments.

Coherences for delta (1.5≤δ<4 Hz), theta (4≤θ<8 Hz), alpha-1 (8≤α-1<10 Hz), alpha-2 (10≤α-2<13 Hz), beta-1 (13≤β-1<15 Hz), beta-2 bands (15≤β-2<25 Hz) and beta-3(gamma-1) bands (25≤β-3<35 Hz) were calculated as the mean coherence values of the three 3-min epochs selected (R, M and Z).

In order to test the spatial homogeneity of EEG coherence, Thatcher et al. (1986) proposed a two-compartmental model of EEG coherence in which different features of

coherence are produced by different length fiber systems. Based on this model, EEG coherence were produced through at least two separate sources (1) the action of short length axonal connection, and (2) the action of long distance connections. Excluding volume conductor contribution, coherence between near electrodes is mainly influenced by short connections, since the axonal density per unit volume of cortex is approximately 10–100 times higher for short-axoned stellate and Martinotti cells than for long-axoned pyramidal cells (Locatelli et al., 1998). On the contrary, the coherence between distant electrodes is mainly due to long axon connections.

In order to evaluate whether the changes were mostly related to the short axonal fibers or to the long ones, coherences were calculated according to Thatcher et al. (1986) by averaged coherence among ‘local’ and ‘far’ recording sites overlying the distribution of different cortico-cortical fibers in order to have information about transmission of the underlying fibers. The average of the coherences between electrodes located in frontal and anterio-temporal regions were used for calculating the local anterior coherence (A: Fp1–F7, Fp2–F8, Fp1–F3, Fp2–F4, Fp1–C3 Fp2–C4, F7–C3, F8–C4, F3–C3 F4–C4, F3–F7 and F4–F8). For the posterior brain region, the local posterior coherence (P) was calculated as the average of coherence values between the electrode pairs: O1–P3, O2–P4, O1–P7, O2–P8, O1–C3, O2–C4, P3–C3, P4–C4, P7–C3, P8–C4, P3–P7 and P4–P8. For posterior to anterior coherence (P-A), we averaged coherences between pairs of electrodes O1–Fp1, O2–Fp2, O1–F3, O2–F4, O1–C3, O2–C4, O1–P3 and O2–P4 and for anterior to posterior coherence(A-P) Fp1–O1, Fp2–O2, Fp1–P3, Fp2–P4, Fp1–C3, Fp2–C4, Fp1–F3 and Fp2–F4, respectively(see Fig. 1(b)). This far coherence is assumed to be the mean coherence transmitted by the superior longitudinal fasciculus in both postero-anterior and antero-posterior directions (Locatelli, et al., 1998). Data from eight homologous right-left (R-L) electrode pairs were also examined as follows: Fp2-Fp1, F8-F7, F4-F3, T8-T7, C4-C3, P8-P7, P4-P3, O2-O1.

Fig. 1 Recording sites for ‘local’ and ‘far’ coherences (a) local anterior and posterior coherences (dotted line) (b) antero-to-posterior and postero-to-anterior directions (dotted line).

5. Statistical analysis

Each EEG data set was divided into 36 non-overlapping, consecutive 5-sec segments. Coherence for each frequency band and for A, P, P-A and A-P connections were calculated from 171 (=18×19/2) pairs composed by 19 electrodes. For each subjects, data was averaged over 36 segments for each session. Means and standard deviations of each state were then calculated. To justify the Normal distribution of the results, we adapted the Kolmogorov-Smirnov Goodness of Fit Test. Statistical differences of coherence were determined using paired Student's t-test for the comparison between baselinestate R versus the meditation states M and Z. Then, results with significant difference (error probability of

P<0.05) are reported below. For each pair of channels, variations between sessions were

evaluated by calculating differences (d-values) between mean coherence values of two sessions, and then normalized by their standard deviation:

R M R Z D C C or C C D n D d = , = − − 2 σ (2) F4 C4 _T8 Fz C3 T7 Pz Cz FP2 F8 FP1 F3 F7 O1 P3 P7 O2 P4 P F4 C4 _T8 Fz C3 T7 Pz Cz FP2 F8 FP1 F3 F7 O1 P3 P7 O2 P4 P (a) (b)

RESULTS

For each meditator, coherences for the different bands showed a similar topographical distribution, without significant side asymmetries, decreasing as distance increased. In order to express all 171 combinations of electrode pairs, values of coherence difference (d-value) are displayed by a series of 19 brain maps. Each map shows the value of the coherence between the selected electrode and the remaining 18 signals. The position of a single map in the series is the position of the electrode taken into account: for example, the first map on the top left series of Fig. 2 shows the increase of theta coherence between Fp1 and the remaining electrodes in the M session compared with R.

All our subjects showed coherence alterations. We found that many alterations are charged to beta band while no apparent modifications are present in slow frequency band as showed in Fig. 2. Coherence value was significantly higher for M and Z states compared with R state over a broad frequency band; the increase was more accentuated for the beta-1, beta-2 and beta-3 coherences.

The lines connecting electrode pairs indicate significantly increased coherences between the given points (Fig. 3). Beta-1 coherence was increased in half the analyzed pairs of electrodes (95pairs, p<0.05) in Z state compared with R state; the increase was more accentuated for the left anterior-to-right posterior coherences and the inter-hemispheric coherences of the anterior to central regions. Similar with beta-1, beta -2 coherence was greater in one-third of electrode pairs (60 pairs, p<0.05) for the Z, more evidently between electrodes over inter-hemispheric frontal- posterior regions and right intra-hemispheric anterior-posterior directions. As for beta-3 band, coherences tended to increase in the anterior and central electrodes in Z state compared with R state. Even if not so marked as the beta changes, there was a coherence decrease presented in Z state compared with R state in the left central regions for the delta band. Examining the difference between M state with R state (as

showed in Fig. 3 (a)), an increase of beta coherence in the same brain area is also present. Tests of spatial homogeneity of EEG coherence were conducted by comparing EEG coherence as a function of different interelectrode distances. The group means for EEG coherence in two meditation sessions and baseline rest are presented in Table 1. Since the coherence decreases as the distance between the two electrodes increase, the values of local and long-distance coherence are in a different range. In comparison with baseline rest, both meditation stages present significant increase in beta coherence in every spatial connection. For low frequency coherences the changes did not reach statistical significance. Higher increases (P<0.01) were more apparent for long distance than for short distance.

-2

0

2

4

delta

theta

beta-1

beta-2

beta-3

d-value

(a) M-R

Fig. 2 Differences of coherence (d-values) for meditation (M) and Zen chakra (Z) stages in compared with baseline rest (R )

Fig. 3 Significant probability mapping showing the comparison between (a) M and R states

-2

0

2

4

delta

theta

beta-1

beta-2

beta-3

d-value

(b) Z-R

and (b) Z and R states. It is evident that Z and M states showed significantly higher synchronization over multiple cortical areas as compared with baseline rest.

Table 1 Regional mean coherence (± standard deviation) in different frequency bands for the free meditation (M), guarding Zen Chakra (Z) and the eye-closed rest (R) sessions (A: anterior; P: posterior; P-A: posterior to anterior; A-P: anterior to posterior; R-L: right to left)

Mean ± SD R M Z M>R Z>R Mean ± SD R M Z M>R Z>R A 0.66±0.06 0.66±0.08 0.68±0.06 － － A 0.62±0.1 0.69±0.1 0.69±0.09 ** ** P 0.62±0.06 0.62±0.06 0.61±0.08 － － P 0.51±0.09 0.55±0.1 0.55±0.1 * * P-A 0.39±0.05 0.4±0.05 0.4±0.07 － － P-A 0.3±0.07 0.36±0.1 0.36±0.1 * * A-P 0.43±0.04 0.43±0.05 0.44±0.03 － * A-P 0.39±0.08 0.46±0.09 0.47±0.09 ** ** δ R-L 0.58±0.05 0.59±0.04 0.58±0.05 － － β-1 R-L 0.45±0.1 0.5±0.1 0.5±0.1 ** ** A 0.71±0.08 0.71±0.08 0.72±0.05 － － A 0.63±0.1 0.69±0.1 0.68±0.1 ** * P 0.55±0.05 0.57±0.07 0.56±0.07 － － P 0.51±0.09 0.55±0.1 0.56±0.09 * ** P-A 0.34±0.03 0.35±0.04 0.34±0.05 － － P-A 0.32±0.07 0.37±0.1 0.38±0.1 * ** A-P 0.45±0.04 0.46±0.05 0.46±0.03 － － A-P 0.42±0.08 0.48±0.1 0.49±0.1 ** ** θ R-L 0.54±0.06 0.56±0.05 0.56±0.05 * * β-2 R-L 0.47±0.1 0.52±0.1 0.53±0.1 ** ** A 0.82±0.1 0.82±0.1 0.82±0.06 － － A 0.59±0.1 0.68±0.2 0.69±0.1 ** ** P 0.52±0.07 0.53±0.08 0.54±0.09 － － P 0.58±0.1 0.63±0.1 0.63±0.1 * * α－1 P-A 0.37±0.1 0.39±0.08 0.4±0.1 － － β-3 P-A 0.4±0.1 0.46±0.2 0.46±0.1 * －

A-P 0.56±0.09 0.57±0.09 0.57±0.08 － － A-P 0.43±0.1 0.53±0.2 0.54±0.1 ** ** R-L 0.62±0.1 0.63±0.1 0.62±0.08 － － R-L 0.48±0.1 0.55±0.2 0.56±0.1 ** **

DISCUSSION

1. Coherence function of different meditative phases

In this experiment, meditation sessions M and Z were characterized by significantly higher beta band coherence over the broad cortical area. The increases are mostly for the beta-1 band, and secondly, beta-2 band. Even if not so marked as the beta-1, beta-3 band coherences also tended to increase in Z state compared with R state.

Our results are different with those in earlier studies. Frontal and central alpha-1 power and frontal-central alpha-1 coherence have been reported during TM practice since 30 years ago (Hebert et al., 2005). Since that report, TM was characterized by significantly higher levels of alpha EEG coherence and has been reported to correlate with improvements in cognitive and emotional parameters such as moral reasoning, emotional stability and anxiety (Hebert et al., 2005). Dissimilar to their findings we have not seen the elevation of coherence reported in the above mentioned paper for the alpha frequency band in meditation. Instead, for low frequency band, coherence value was close to that obtained for the wakeful rest state.

We observed an increase in beta coherence in comparison with wakefulness for most electrode combinations. Higher levels beta power and coherence indicates cortical areas involved in task processing and functional coupling (Thatcher et al., 1986) between brain regions. Also, transient synchronization of neuronal activity seems to be a key mechanism in the binding of anatomically distributed feature processing into coherent perceptual objects, where it is often associated with β or γ oscillations (Palva & Palva, 2007). Because Z state involves internal attention inside the third ventricle, enhancement of global, fast-band coherence may indicate the synchronization of massive distributive neuralassemblies via the focusing process during Z state.

To make a general survey of topographical characteristics for band wave activities, topographic maps of relative power and coherence were made different frequency bands corresponding to EEG of R, M and Z stages. Fig. 4 shows average patterns across all meditators of topographical maps of relative power (upper row of each panel) and coherence (lower row of each panel). The color bars on the right side of this panel indicate the level of relative band power while the color bar on the bottom side indicates the coherence strength. To be simplification, only 35 electrode pairs (~20%) with the highest coherence were showed. Darker lines of topographic maps of coherence show higher coherence level with lighter lines show the lower level. In these stages (R, M, Z), there were no the remarkable changes of topographic characteristics in alpha-2 activities.

Topographic maps of theta and alpha-1 band demonstrated that the relative power increased as the function of meditation stages in the anterior-central areas of scalp but the changes of the coherence values are not significant. On the other hand, Topographic maps of coherence in beta-1 and beta-2 activities clearly demonstrated that the synchronous component in the anterior-central areas of scalp appeared to correspond with increasing relative power. Relative power of beta-3 band increased from occipital region to frontal region and coherence values were also high at those areas.

In general, increased alpha power over central and frontal cortices indicates cortical areas at rest or ‘‘cortical idling’’ and could indicate decreased motor and executive processing during TM practice (Travis, 2001). Inspired by these results we thus propose that the meditators are more relaxed during M state with mental quiescence but full inner wakefulness. Even though power of fast rhythms is not so significantly different between two meditative states due to the low amplitude, the coherence estimate is. Coherence may be more sensitive than amplitude to distinguish Z state from the M and baseline R states. As a result, the increase of beta coherence was more relevant in Z state, and the increase of low frequency power was more relevant in the M state.

Fig. 4 Topographic maps of relative power (upper row of each panel) and coherence (lower row of each panel) for different frequency bands corresponding to EEG during R, M and Z stages. To be simplification, only 35 electrode pairs with the highest coherences were showed.

REFERENCES

Chang, K. M., Lo, P. C., 2005, "Meditation EEG Interpretation Based on Novel Fuzzy-Merging Strategies and Wavelet Features." Journal of Biomedical

Engineering-Applications, Basis & Communications, Vol. 17, pp.167-175

Hebert, R., Lehmann, D., Tan, G., Travis, F. and Arenander, A., 2005, "Enhanced EEG alpha time-domain phase synchrony during Transcendental Meditation: Implications for cortical

Pr Pr

Pr Pr

integration theory," Signal Processing, Vol. 85, pp. 2213-2232

Locatelli, T., Cursi, M., Liberati, D., Franceschi, M. and Comi, G., 1998, "EEG coherence in Alzheimer's disease," Electroencephalography and Clinical Neurophysiology, Vol. 106, pp. 229-237

Nunez, P.L., Silberstein, R.B., Shi, Z., Carpenter, M.R., Srinivasan, R., Tucker, D.M., Doran, S.M., Cadusch, P.J. and Wijesinghe, R.S., 1999. "EEG coherency. II. Experimental comparisons of multiple measures." Clin Neurophysiol, Vol. 110, pp. 469–486.

Palva, S. and Palva, J. M., 2007, "New vistas for a-frequency band oscillations," Trends in

Neurosciences, Vol. 30, pp. 150-158

Thatcher, R.W., Krause, P.J. and Hrybyk, M., 1986, "Cortico-cortical associations and EEG coherence: a two-compartmental model." Electroenceph. clin. Neurophysiol. Vol. 64, pp. 123–143

Travis, F., 2001, "Autonomic and EEG patterns distinguish transcending from other experiences during transcendental meditation practice." International Journal of

Psychophysiology, Vol. 42, pp. 1–9.

Welch, P.D., 1967, "The Use of Fast Fourier Transform for the Estimation of Power Spectra: A Method Based on Time Averaging Over Short, Modified Periodograms." IEEE Trans.