禪坐中心血管系統及中樞神經系統電生理現象之研究

國 立 交 通 大 學

電機與控制工程學系

博士論文

禪坐中心血管系統及中樞神經系統電生

理現象之研究

Research on Cardiovascular System and

CNS Electrophysiological Phenomena

under Zen Meditation

研究生：劉權毅

指導教授：羅佩禎 博士

中華民國九十六年十二月

禪坐中心血管系統及中樞神經系統電生理現象之研究

Research on Cardiovascular System and CNS

Electrophysiological Phenomena under Zen Meditation

研 究 生：劉權毅

Student：Chuan-Yi Liu

指導教授：羅佩禎

Advisor：Pei-Chen Lo

國 立 交 通 大 學

電機與控制工程學系

博士論文

A Dissertation

Submitted to Department of Electrical and Control Engineering

College of Electrical and Computer Engineering

National Chiao Tung University

In Partial Fulfillment of the Requirements

For the Degree of

Doctor of Philosophy

in

Electrical and Control Engineering

December 2007

獻給我的父母

劉正信先生與游燕墉女士

For My Dear Parents

Cheng-Hsin Liu and Yen-Yung Yu

禪坐中心血管系統及中樞神經系統電生理現象之研究

研 究 生：劉權毅

指導教授：羅佩禎 博士

國立交通大學 電機與控制工程研究所

摘要

本篇論文在研究禪坐對於心血管系統，以及對於中樞神經系統的影響。在評 估心血管系統的效應中，我們使用了血液壓力波的四種參數來量化禪坐對於血管 系統的效果，這四種參數包括了 T 波的上昇斜率，以及 T 波、V3 波谷以及 D 波 的正規化的高度。我們的實驗結果顯示，禪坐會加強心臟的搏出能力，得到較佳 的動脈順應性，降低血管的阻塞，並強化動脈彈性以及主動脈瓣膜的功能。這些 觀察結果可推論出禪坐的確可以促進心血管系統的相關特性。 禪坐者在禪坐過程中，常會體驗到有內在光芒的出現，為了研究此一普遍現 象，我們設計了閃光視覺誘發電位實驗。在禪坐(放鬆)的前中後，我們對實驗組 (控制組)受測者施以閃光刺激並蒐集視覺誘發電位。我們的實驗結果顯示在禪坐 的過程中，後腦(Oz)位置上較晚的視覺誘發電位成分 N3-P3 以及 P3-N4，其振幅 會較禪坐前降低，但是對於控制組(放鬆)而言，這些成分的振幅卻是增大的。此 外，這些成分在中腦與前腦的振幅，在禪坐過程中是增加的，在休息中則是減少 的。而某些成分的潛時(latency)在休息中是延長的，而禪坐中則是沒有太大的

質層的特定效應，和單純閉眼休息造成的效果是不一樣的。 在第四章中，我們探討了禪坐腦電波對於 alpha 波的空間聚集效果，並有了 初步的結果。我們分別對於實驗組(禪坐)與控制組(閉眼放鬆)錄製腦電波，並使 用小波分析方法對多頻道腦電波進行萃取 alpha 波的工作。我們再對正規化的 alpha 能量向量使用模糊分類演算法進行分類，以探討不同的空間分佈特性。分 類過程中，我們對各群中心的模糊關係值向量進行相似度比對，以找出最佳的分 類數。我們的結果顯示，前腦 alpha 活動發生的機率與禪坐的階段有關。從結果 中可證實，禪坐過程中會產生三種空間-時間變化的分佈類型，而這是由於禪坐 過程中的三個階段所產生的。此外，這兩組的後腦 alpha 活動也有不同的趨勢， 控制組在禪坐過程中，後腦 alpha 的能量隨著時間呈現遞減的現象，但是實驗組 則的後腦 alpha 一直呈現低於平常準位的現象。這些研究結果顯示，禪坐的過程 大致上會分成三個階段，而各個階段對於 alpha 調變的機制有所差異。 在第三與第四章中，我們觀察到禪坐過程中，前額 alpha 有增加的趨勢，且 視覺誘發電位與閉眼休息的情形下，也有不一樣的變化；因此我們設計了一種新 的實驗，使受測者在相同的背景腦電波的條件－前額出現由 alpha 波主導的情況 下，我們才對受測者施予閃光刺激。我們觀察了實驗組(禪坐者)與控制組(一般 未練習禪坐者)兩者之間視覺誘發電位的差異，結果顯示，在禪坐中時，Cz 及 Fz 的位置上 P1-N2 and N2-P2 成分的振幅明顯增大，而控制組在休息時這些成 份則會降低。因此，我們推測禪坐對於主要視覺皮質層及相關區域會有特定作用。

Research on Cardiovascular System and CNS

Electrophysiological Phenomena under Zen Meditation

Student：Chuan-Yi

Liu

Advisor：Dr. Pei-Chen Lo

Institute of Electrical and Control Engineering

National Chiao Tung University

Abstract

This dissertation reports the effects of Zen meditation on cardiovascular system

and CNS electrophysiological behaviors. In the aspect of cardiovascular study, we evaluated the effects of meditation on cardiovascular system based on four parameters

derived from blood pressure wave (BPW) that included the rising slope, normalized height of T wave, normalized height of valley, and normalized height of D wave.

The results showed that Zen meditation could lead to better heart ejection ability and aorta compliance, better arterial elasticity and aortic valve function, as well as the

decreasing peripheral resistance of blood vessels. The observation allows us to infer that Zen meditation effectively improves functional characteristics of the

cardiovascular system.

3

V

To investigate the common experience of meditation practitioners - perception of

inner light during meditation, we designed the flash visual evoked potentials (F-VEPs) experiments. Flash stimuli were applied before, during and after meditation- /

relaxation-session in experimental / control subjects. Our results show that amplitudes of late latency components N3-P3 and P3-N4 at Oz decrease during meditation in the

experimental group, whereas they increase in the control group. Both Cz and Fz amplitudes increase during meditation, yet decrease during relaxation for the control

group. The latencies of some components increase under relaxation in control group, yet little variation (except P2) is observed in the meditators. According to our findings,

Zen meditation induces particular effects on the visual nervous system and cortex that are distinguishable from those observed for normal relaxation.

In chapter 4, we report our preliminary results of investigating the spatial focalization of Zen-meditation EEG (electroencephalograph) in alpha band (8-13 Hz). For comparison, the study involved two groups of subjects, practitioners (experimental group) and non-practitioners (control group). Wavelet analysis was applied to multi-channel EEG signals to extract the alpha rhythm. Normalized alpha-power vectors were classified by Fuzzy C-means based algorithm to explore various brain spatial characteristics. Number of clusters was determined by correlation coefficients of the membership-value vectors of each cluster center. Our results show that, in the experimental group, the incidence of frontal alpha activity varies in accordance with the meditation stage. And the results demonstrated three different spatiotemporal modules consisting with three distinctive meditation stages normally recognized by meditation practitioners. The parietal alpha activity in two groups decreased in different ways. Particularly, monotonic decline was observed in the control group. The phenomenon might imply various mechanisms employed by meditation and relaxation in modulating parietal alpha.

In the research of visual perception under meditation based on alpha-dependent

F-VEPs, we designed a strategy to record the F-VEPs under the consistent condition of background EEG – the emergence of dominant frontal alpha activity. According

our preliminary results, the frontal alpha rhythms were more active during meditation. This event was thus used as the reference of meditation stage to trigger the flash

stimuli. Based on the experimental protocol proposed, considerable differences between experimental and control groups were observed. In sum, amplitudes of

the F-VEPs of control group at rest. We thus suggest that Zen meditation results in acute response on primary visual cortex and the associated parts.

誌 謝

這段追求真理的時光，是我最快樂、最重要的歲月。 首先感謝指導教授羅佩禎老師，這幾年老師不只在專業領域上盡心盡力的教 導，更在人生真理上給予我們更高更遠的視野；感謝老師始終秉持原則不斷挑 剔、鞭策、引導我們，在這條孤獨的研究道路上，給予我信心與勇氣繼續往下走。 感謝口試委員楊谷洋、邱俊誠、謝仁俊、張剛鳴老師，老師們對於論文的不 吝指導，讓我在論文研究上獲益匪淺。 感謝實驗室的學長姐：清泉、政勳、瑄詠，你們的指導與經驗，指引了我的 研究方向，也導正了我許多的迷惘與疑惑。 特別感謝憲正與適達兩位同學，有你們兩位的互相砥礪，互相打氣，互相扶 持，我才能在這條路上走下去；能有緣在這段時光與你們同行，是我最大的福氣； 雖然往後大家各奔東西，但我會一直記得你們坐在我身旁，一起努力的景象。 感謝這幾年不斷麻煩我，騷擾我，幫助我的各位學弟：小波波、進忠、啟宏、 仁隆、維廷、哲賢、槍哥、強哥、清文、政恩、昶毅、恩榮、敬達、小胖胖、Bono、 宏彥，有你們的歡樂與活力，讓我的博班生活充滿了歡笑與光彩。 感謝室友阿昌這幾年來與我互相哈拉，也感謝他教我這麼多攝影的知識技 術，雖然有很多我聽不懂。 感謝當初碩班的同學焦仕揚、威助、治亨，雖然只短短同學兩年，但是對我 往後博班的影響與助益卻相當深遠。 最重要的是，感謝我的父母及兄姐，有你們的支持，我才能毫無後顧的投入 這條研究的道路，才有可能完成這個人生的夢想；謝謝你們無怨無悔的付出，我 永遠愛你們。 感謝大家的愛與支持，給予我源源不斷的溫暖，謝謝大家！

Contents

摘要---i Abstract --- iii 致謝--- vi Contents--- vii List of Tables--- x List of Figures--- xi 1. Introduction --- 1 1.1 Background --- 3

1.2 Aim of this work --- 5

1.3 Organization of the Dissertation --- 8

2. Variation Analysis of Sphygmogram to Assess the Cardiovascular System under Zen-Meditation ---10

2.1 Background and Motivation ---10

2.2 Methods for BPW Analysis---12

2.2.1 Mechanism and Recording Procedure ---12

2.2.2 BPW Parameters ---18

2.3 Experiment and Results ---20

2.3.1 BPW signal Processing ---20

2.3.2 Subjects and Recording Paradigms---22

2.4 BPW Before and After Meditation---23

3. F-VEPs in Zen-Meditation ---29

3.1 Background and Motivation ---29

3.2 Experimental setup and Procedure---30

3.3 Results and Inter-Group Comparison---31

4. Spatial Focalization of Zen-Meditation Brain Based EEG---40

4.1 Background and Motivation ---40

4.2 Proposed Schemes ---43

4.2.1 Wavelet Transform---43

4.2.2 Alpha Detection ---43

4.2.3 Fuzzy C-means---45

4.3 Subjects and Recording Setup---50

4.4 Results ---51

5. Investigation of alpha-dependent F-VEPs under Zen-Meditation---61

5.1 Motivation of this research ---62

5.2 Systems and approach ---64

5.2.1 Online alpha-rhythm detection---65

5.2.2 Simulation ---67

5.2.3 Off-line alpha detection ---69

5.2.4 F-VEPs ---71

5.3 Experimental Setup and Protocol ---73

5.3.1 Subjects ---73

5.3.2 Apparatus ---73

5.3.3 Experimental paradigms ---75

6. Discussion and Conclusion---83

6.1 Variation Analysis of Sphygmogram to Assess the Cardiovascular System under Zen Meditation---83

6.2 F-VEPs in Zen-Meditation---85

6.3 Spatial Focalization of Zen-Meditation Brain Based EEG ---86

6.4 Investigation of alpha-dependent F-VEPs under Zen-Meditation ---88

6.5 Discussion---89

6.6 Future Work---91

Bibliography ---92

Appendix --- 105

List of Tables

Table 2-1: The statistical results of four parameters and their variation percentages.28

Table 3-1: Average ratios of peak latencies ---36

Table 3-2: Average ratios of peak amplitudes---38

Table 4-1: The correlation coefficients (4 clusters) ---47

Table 4-2: The correlation coefficients (3 clusters) ---48

Table 4-3: The Euclidean distances between cluster centers---51

Table 4-4: The standard deviation of the distances of each sample to its centers ---51

Table 5-1: Locations of poles of the simulated signal ---69

List of Figures

Figure 2-1: Prototype of a normal blood pressure waveform--- 13

Figure 2-2 (a): When the ventricle contracts, the semilunar valve opens. Blood ejects into the aorta and arteries and makes them expanded. --- 14

Figure 2-2 (b): Isovolumic ventricular relaxation makes the semilunar valve shut. --- 15

Figure 2-3: The acquiring instrument of blood pressure waveform--- 17

Figure 2-4: The measuring position on the wrist --- 18

Figure 2-5: The linear baseline drift in the BPW and the result after removing it --- 21

Figure 2-6: The before-meditation and after-meditation BPW for an experimental subject --- 24

Figure 2-7: The ten-second typical BPW of one meditation practitioner --- 25

Figure 3-1: The F-VEP on (a)Oz (b)Cz (c)Fz of a meditator --- 34

Figure 3-2: The F-VEPs on (a)Oz (b)Cz (c)Fz of a control subject--- 35

Figure 4-1: A section of 5-sec EEG. The numbers are the alpha-power percentages --- 45

Figure 4-2: The results of 4 clusters--- 47

Figure 4-3: The results of 4 clusters--- 48

Figure 4-4: Flowchart of the proposed algorithm --- 49

Figure 4-5: Selected samples (three rows) and the center of Cluster #1.--- 52

Figure 4-6: Selected samples (three rows) and the center of Cluster #2 --- 53

Figure 4-7: Selected samples (three rows) and the center of Cluster #3 --- 53

Figure 4-8: The color chart of alpha distribution --- 55

Figure 4-9: The locations of 30 recording electrodes and their respective region --- 55

Figure 4-10: The incidence of frontal alpha of both groups --- 57

Figure 4-11: The incidence of central alpha of both groups --- 58

Figure 5-1: Tree structural filter bank for the Subband-AR EEG Classifier. --- 65

Figure 5-2: Classification result of the simulated signal --- 68

Figure 5-3: Result of α detection for real EEG signal --- 70

Figure 5-4: Profile of F-VEPs on (a) Fz, (b) Cz, and (c) Oz--- 72

Figure 5-5: Experimental setup for α-dependent F-VEP recording--- 74

Figure 5-6: Scheduling of the F-VEP recording procedure --- 76

Figure 5-7: Display format of selected channels (Fz, Cz, Pz, and Oz) for α-dependent F-VEP recording --- 77

Figure 5-8: The α-dependent F-VEPs of one meditator recorded on (a) Fz, (b) Cz, and (c) Oz --- 80

Figure 5-9: Variations of N2-P2 amplitudes at Cz and Fz--- 81

Figure A-1: The scheme diagram of the low-pass and high-pass filter --- 106

Figure A-2: The wavelet decomposition tree --- 107

Figure A-3: The process of decomposition and reconstruction--- 107

Figure A-4: The process of reconstructing the low frequency component --- 108

Chapter 1

Introduction

Due to the therapeutic effectiveness, the new area CAM (complementary and

alternative medicine) has drawn the attention of researchers and medical

professionals in the past decades. Researches in biomedical engineering and life

sciences should lay more stress on promoting the human health, in both the

physiological and mental aspects. As a matter of fact, scientists of the West have been

reporting substantial findings of the effectiveness of meditation practice in CAM not

only on improving the physiological and mental health but on treating a number of

diseases. Accordingly, our research laboratory (Biomedical Signal Research Lab) has

been devoted to the study of Zen-Buddhist meditation since 1998. We investigate,

from the viewpoint of biomedical engineering, phenomena of the human life system

under the orthodox Zen meditation practice. We study the time-varying characteristics

and dynamic mechanism during meditation course in order to further establish the

correlation among different electrophysiological signals and parameters.

Meditation, classified as the category of mind-body intervention in

researches have been devoted to the study of meditation process and phenomena,

mostly in the physiological and psychological aspects. Among various meditation

techniques, we focus on Zen meditation that has been becoming popular during the

past decades.

Since 1998, we have been investigating the Zen-meditation

electroencephalography (EEG) and other physiological parameters in multi-faceted

views, and abundant data and results have been accumulated and reported. This

dissertation is mainly devoted to the study of Zen-meditation effects on

cardiovascular system and brain electrophysiological behaviors based on the

observation of blood pressure wave (BPW), EEG and flash evoked potentials

1.1 Background

Scientific exploration has corroborated the effectiveness of meditation practice on

the health promotion which includes regulation of the hormone-level and blood

pressure, moderation of stress and anxiety, reduction of chronic pain, etc (Dillbeck

and Orme-Johnson, 1987; Jones, 2001; Walton et al., 1995; Barnes et al., 2001;

Lindberg, 2005; Shetty, 2005). According to the experienced practitioners, meditation

facilitates a greater sense of calmness, empathy, and compassion. As Western medical

practitioners begin to understand the role of mind in health and disease, there has been

more interest in both employing meditation in medicine and exploring brain

dynamical phenomena during meditation. Electroencephalogram (EEG) thus becomes

an important tool to monitor the meditation process (Niedermeyer and Lopez da Silva,

1999).

A thorough review of researches on meditation EEG can be found in (Cahn and

Polich, 2006). Although much research has been devoted to investigating the

physiological effects of meditation, many phenomena still have not yet been

completely understood by people in CAM or in mainstream West medicine.

According to our long-term interactions with the experienced practitioners for several

years, the Zen meditation process involves experience of transcending various

would first attenuate their physical and mental sensors via particular mind-focusing

technique, leave off the message transmission from outside world, and keep

subconsciousness tranquil during meditation. Moreover, meditation practitioners often

experience unusual perceptions, for example, loss or distortion of space and time

perceptions, sensation of aureola-surroundings, etc. Especially, in the deeper

meditation state, many meditation practitioners have experienced the perception of

inner light (Lo et al., 2003). However, quite few studies were focused on this unique

phenomenon reported by meditation practitioners. Accordingly, we aimed to

investigate meditation effects on the visual neuronal pathway by quantifying the flash

visual evoked potentials (F-VEPs). Furthermore, we investigated the relation between

F-VEPs and frontal alpha waves often observed during meditation.

In addition, most studies about meditation effects on cardiovascular system

emphasize the variation of blood-pressure range or electrocardiogram (ECG) signal,

but few on the BPW (blood pressure wave) characteristics. Having been widely

adopted in traditional Chinese medicine, BPW provides a more pronounced medium

for characterizing Zen-meditation phenomena based on the concept of Qi-energy. This

dissertation also presents a quantitative approach for investigating the variation of

1.2 Aim of this work

The main aim of this dissertation is to investigate the effects of Zen meditation on

the cardiovascular system and brain electrophysiological behaviors. To attain this aim,

we designed and carried out four experiments, for each presumed meditation

phenomenon, to probe and further validate the results.

A number of literatures have reported the meditation effects on cardiovascular

system. In particular, better control of blood pressure and modulation of heart rate

variation (HRV) have drawn considerable attention of researchers (Barnes et al.,

2004). They demonstrated that meditation might lower the blood pressure and

therefore benefit people suffering from hypertension. To the best of our knowledge,

the mechanisms digging into cardiovascular system behaviors have not been

discussed.

In recent years, blood pressure wave, reflecting the meridian energy in TCB

(traditional Chinese medicine) for thousand years, has been also employed in clinical

diagnosis in conventional West medicine. The parameters of BPW have been used to

investigate the characteristics of blood vessels of cardiac patients and normal subjects

(Fey, 2003). Our previous study on HRV under Zen meditation revealed an apparent

benefit to ANS (autonomous nervous system) and, hypothetically, to the

meditation on cardiovascular system based on BPW analysis.

During the past decades, numerous researches about meditation EEG have been

reported (Cahn and Polich, 2006). A few of them dealt with the characteristics of

evoked potentials affected by meditation. Evoked potentials are the responses of

central nervous system (CNS) to the external stimuli, including sounds (brainstem

auditory evoked potentials, BAEP), light or pattern variations (visual evoked

potentials, VEP), or electrical stimulation of peripheral nerves (somatosensory evoked

potentials, SSEP). Each unique evoked potential characterizes the neuronal pathway

corresponding to the control of particular sensory organ. According to the narration

of Zen-meditation practitioners in our experiments, inner-light perception (Lo et al.,

2003) is one of the common experiences in this group of subjects. This prompted us to

study the meditation effects on visual neural pathway based on quantitative analysis

of visual evoked potentials.

According to the researches of meditation EEG, variation of frontal alpha has been

identified to be an exclusive brain electrophysiological phenomenon during or after

the meditation course. In our study, the meditation session lasted 30 minutes or more.

Note that EEG is non-stationary at the level of second. We developed a scheme, based

on wavelet theory and fuzzy c-means, to investigate the spatial-temporal distribution

From the F-VEPs and spatiotemporal alpha studies described above, we noticed

the issue of F-VEP variations due to background EEG. To make the experiment more

conscientious, we proposed a hypothesis that F-VEP should be conducted under the

same background EEG. As a consequence, a real-time alpha detection scheme was

required. In this study, we collected F-VEPs upon the event that alpha wave presented

1.3 Organization of the dissertation

This dissertation is composed of six chapters. In chapter 1, we introduce the

background and the aim of our research. In addition, motivation and background of

designing four studies are described. Chapter 2 reports the results and our findings of

BPW research. The beginning of this chapter is devoted to the BPW mechanism, the

experimental setup, and the signal processing method. The results and statistical

analysis are presented at the end of the chapter.

Chapter 3 is focused on the F-VEP study. The first part includes the review of

researches related to our work and the motivation of conducting this study. Illustration

of experimental procedure follows. Finally the inter-session and inter-group

comparisons are conducted.

In chapter 4, we proposed a method to monitor the spatio-temporal distribution of

alpha waves using wavelet and fuzzy c-means. Finally, inter-group difference is

justified by statistical analysis.

Chapter 5 describes the scheme for real-time alpha-rhythm detection employed in

the F-VEP study. This study was attempted to explore the meditation effects on visual

perception.

The last chapter summarizes the findings of four studies, that is, BPW, F-VEPs,

Chapter 2

Variation analysis of

sphygmogram to assess the

cardiovascular system under

Zen meditation

In this chapter, we studied how meditation affects the characteristics of

cardiovascular system, mainly based on the blood pressure waveform (BPW). Four

parameters derived from the BPW include the rising slope (

1 1 t h ), normalized height of T wave ( 1 3 h h

), normalized height of V₃ valley (

1 4

h h

), and normalized height of D wave

(

1 5

h h

), where t1 and hi, i = 1, …, 5 are quantitative features of the BPW waveform

pattern.

2.1 Background and motivation

The blood pressure waveform (BPW) of the systemic arterial tree is an important

systole and diastole of the heart and conveys such information as the blood ejection

ability of the heart, the elasticity of the artery wall, the peripheral resistance, etc.

(Milnor, 1989). In examinations of the clinical value of BPW, Han explored (Han,

2000) possible biophysical and pathological mechanisms of BPW from the viewpoint

of hemodynamics. Research showed that BPW analysis is a highly reproducible

method and easy to apply to clinical studies. This measure provides important

information about arterial stiffness and cardio-vascular interactions (Wilkinson, 1998;

O'Rourke1, 2001). Abnormality in the blood pressure waveform is linked to various

physiological or pathological states such as aging and hypertension (Cohn et al., 1995;

Mcveigh et al., 1999). Actually, the blood pressure waveform of radial artery detected

at the wrist is the sphygmographical signal used in Traditional Chinese Medicine

(TCM) (Tan, 2004). According to theory of the sphygmographical signal, the TCM

clinician can identify the status of the human body and treat the patient.

As more clinical evidence supported the benefits of meditation for health, about

fifty years ago researchers began investigating the physiological phenomena of the

human body under meditation. Dillbeck et al. (1987) compared the physiological

differences in two groups of subjects, one under transcendental meditation and the

other at rest. Schneider et al. (1995) found that the training of transcendental

hypertensive persons (Barnes et al., 2004; Alexander et al., 1996; Castillo-Richmond et

al., 2000). Meditation hereafter became a feasible method to improve the hypertension.

Hankey compared Tibetan Buddhist meditation with Transcendental Meditation

(Hankey, 2006). He summarized how practicing different meditation techniques

influenced hypertension and other physiological changes. Barnes et al. (1999) found

that, under meditation, total peripheral resistance decreased, and they suggested that

was why meditation could decrease or control hypertension. To investigate the

meditation effects on the cardiovascular system, here we evaluated the variations in

blood pressure waveform before and after meditation.

We measured the blood pressure waveforms of twenty Zen-meditation

practitioners and twenty normal, healthy subjects in the same age range as the

practitioners. According to the clinical experience of TCM professionals, we designed

a set of parameters that quantify the waveform patterns of BPW.

2.2 Methods for BPW analysis

2.2.1 Mechanism and recording procedure

The BPW prototype of a healthy subject is shown in Fig. 2-1. The heart pumping

mechanism correlating with BPW is illustrated as follows (see Fig. 2-2). The ejection

called the Percussion wave (P wave). The height of P wave, , and the fast ejection

time of the left ventricle, , are related to the ejection ability of heart and the

compliance index of aorta. We define the rising slope of P wave as

1 h 1 t 1 1 t h . A larger slope

indicates a better performance of the heart ejection function and aorta compliance

(Fey, 2003). Thus it is used as a quantitative feature to evaluate the cardiovascular

system. P wave T wave D wave 1

h

3

h

5

h

4

h

3 V 1 V 1

t

Figure 2-2 (a): When the ventricle contracts, the semilunar valve opens. Blood ejects

into the aorta and arteries and makes them expanded. At the same time, the pressure

Figure 2-2 (b): Isovolumic ventricular relaxation makes the semilunar valve shut.

Elastic recoil of arteries sends blood forward into the rest of circulatory system. This

The second peak, called the Tidal wave (T wave), appears when blood hits the

artery wall and rebounds. As a result, T wave is manifest if the artery possesses

excellent elasticity that reflects low peripheral resistance of the circulatory system. On

the other hand, an artery with a stiff wall makes the T wave propagate fast according

to the Moens-Korteweg equation of wave velocity (Nichols et al., 1990; Khir et al.,

2002). Accordingly, the T wave will merge with the P wave, which results in a wider P

wave. The second parameter,

1 3

h h

, where represents the height of T wave, is

utilized to measure the effect of T wave. We thus expect a large

3 h 1 3 h h for an arterial

system with better elasticity. The valley height reveals the level of peripheral

resistance (Fey, 2003; Milor, 1989). As the peripheral resistance increases (decreases),

parameter increases (decreases) as well. The normalized parameter,

4 h 4 h 1 4 h h , is

employed to measure the drift of peripheral resistance. Finally, when the aortic valve

is closed, the Dicrotic wave (D wave) is generated. is the magnitude of D wave

and the normalized parameter,

5 h 1 5 h h

, represents the effect of D wave on arterial system.

will decrease due to a stiff aorta or aortic regurgitation.

5

h

Fig. 2-3 displays the instrument for recording BPW from the wrist with a

piezoelectric sensor. The sensor was manufactured by Skylark Company, with 3dB

cutoff frequency at 10 KHz. The output is linearly dependent of the input when BP is

upper arm and desk is about 120°. Pressure of the piezoelectric sensor on the wrist was adjusted for sufficient sensitivity of the BPW activity. The range of amplitude

depends on the measuring position. In this study, we measured the BPW at the ‘chun’

position (Fig. 2-4). A well experienced TCM (Traditional Chinese Medicine) expert

assisted us in the experimental setup and recording. He searched the “chun” position

by palpating the wrist of the subject. The piezoelectric sensor was then attached to the

proper position carefully identified by the expert. The pressure was adjusted until the

output amplitude reached maximum, from which the amplitude recorded was in the

range of 15 to 27 mmHg (Fey, 2003).

Figure 2-4: The measuring position on the wrist.

2.2.2 BPW parameters

The peak or valley positions in Figure 2-1 demonstrated by crosses are

determined by the Matlab program for searching the local maximum or minimum. We

first identify the P wave that is the most distinctive one. The starting point V1 of the

whole period is about 0.1 sec prior to the P wave. The next step is to find the local

minimum V3 at about 0.35 sec afterward from V1. Then the T wave between P wave

and V3 is ready to be extracted. D wave comes after V3, that is identified by finding

the null of waveform differentiation between V3 and the end of the BPW period. The

time values of these specific points depend on the blood speed and the conditions of

blood vessels. We referred to the paper by Xie (Xie et al., 2000) to check the standard

time values of these parameters.

In this research, the above four parameters,

1 1 t h , 1 3 h h , 1 4 h h and 1 5 h h are used to

assess the status of cardiovascular system and corresponding parameters are defined

as follows:

1 1

1 1

h = Peak height of P wave – Valley height of V (2.1)

3

h = Peak height of T wave – Valley height of V₁ (2.2)

4

h = Valley height of V₃ – Valley height of (2.3) V₁

5

h = Peak height of D wave – Valley height of V₃ (2.4)

t = Time of V – Time of Peak of P wave (2.5)

Variation percentage = after meditation before meditation 100%

before meditation η η η − − − − × , (2.6)

where η denotes one of the four parameters,

1 5 1 4 1 3 1 1_{, ,} h h or h h h h t h

. Note that, due to the

variations in recording system characteristics and physiological conditions of each

subject, heights h3-h5 were normalized to within the range of 0 to 1. Normalization

ensures an even comparison of BPW features in various recording sections. To avoid

null denominator caused by the disappearance of before-meditation T wave (in the

case, h3=0), the variation percentage of

1 3

h h

is defined to be zero if T wave is absent

before and after meditation. On the other hand, if T wave exists only after meditation,

the variation percentage of

1 3

h h

2.3 Experiment and results

2.3.1 BPW signal processing

The blood pressure waveforms of radial artery non-invasively detected at wrist

by piezo-electric transducer were recorded for 10 seconds and digitized at a sampling

rate of 100 Hz. To reduce high frequency noise, a low-pass filter is designed with a

3dB cutoff frequency of 50 Hz. Because null baseline was required in the analysis, we

removed the mean value of each BPW period in pre-processing. Accordingly, negative

waveforms might appear. Note that non-constant, linear baseline drift sometimes

interfered in the BPW (Fig. 2-5). The sources of interference include the body

movement, heavy respiration, etc. The resulting artifacts might be false amplitude or

position shift of peaks/valleys. Although removal of the baseline drift may help, we

found it necessary to further correct the interference patterns. In that case, baseline

Figure 2-5: The linear baseline drift in the BPW (upper) and the result after removing

it (lower).

The BPW recording instrument can only continuously trace for up to 10 seconds.

BPW is a quasi-periodic signal with 10 to 16 complete cycles in a 10-second term.

Here we averaged out the 10-16 cycles to get the final BPW for further quantification

of embedded features. We thus can obtain triple to quadruple SNR (signal-to-noise

ratio) that is linearly dependent of the square root of the number of measurements.

2.3.2 Subjects and recording paradigms

The participants were divided into two groups - 20 meditators and 20 normal,

healthy people without any experience in meditation. In the experimental group, 13

females and 7 males with a mean age of 26.6±2.2 years participated. Their

experiences in Zen-Buddhist practice span 6.9±3.3 years. The control group

comprised 9 females and 11 males with a mean age of 25.2±1.8 years. All the

meditation practitioners learned Zen-Buddhist meditation in the Taiwan Zen-Buddhist

Association. Only experienced practitioners with more than 3 years of meditation

experience were invited. The controls were graduate students of National Chiao-Tung

University. No subjects had any cardiovascular disease in their medical histories.

Participants sat in an isolated space during the experiments. Blood pressure

waves were recorded before and after the 40-minute main session (meditation or

relaxation). In the main session, experimental group practiced Zen meditation, while

2.4 BPW before and after meditation

Fig. 2-6 illustrates an example of the blood pressure waveform of meditator. The

solid curve shows the before-meditation BPW, and the dashed one shows that after

meditation. In Fig. 2-6, after-meditation P wave rises more steeply than the

before-meditation one. For this particular subject, poor arterial elasticity makes the

before-meditation T wave occur earlier and merge with P wave, resulting in a

broadened P wave. In after-meditation BPW, the T wave becomes evident and

distinguishable from the P wave, inferring that enhanced arterial elasticity is a

consequence of meditation practice. Moreover, the V3 valley descends (i.e., h4

decreases) after meditation, indicating a decrease in peripheral resistance. The D wave

magnitude is also strengthened after meditation. A typical example of 10-second BPW

is shown in Fig. 2-7. Note that the pulse rate is slower after meditation, resulting in a

phase difference between the two waves. In sum, the after-meditation BPW reflects a

more robust cardiovascular system that could have been tuned up by the Zen

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 -5 0 5 10 15 20 Time (Sec) B loo d P re s s ur e W a v e ( m m H g) Before meditation After meditation

Figure 2-6: The before-meditation (solid curve) and after-meditation (dashed curve)

1 2 3 4 5 6 7 8 9 10 -15 -10 -5 0 5 10 15 20 25 Time (Sec) B lood P res s ur e W av e ( m m H g) Before meditation After meditation

2.5 Inter-group comparison

Table 2-1 shows the results of the four parameters measured in both groups. The

variation percentage of each parameter in each group is calculated by the formula

(2.6). The P-values in Table 1 are evaluated using t-test, which is used to check

whether the variation percentages show statistical differences between the groups. In

this preliminary investigation, we concentrated on intra-subject differences between

various experimental sections because the inter-subject variations in BPW were too

complicated to manipulate.

In comparison with the control group, Zen-meditation practitioners have higher

ranges of variation percentages in all four parameters. The performance details of

each parameter are as follows: When considering the rising slope of P wave,

1 1

t h

,

that reflects the ejection ability of left ventricle or aorta elasticity, the experimental

group had a mean increase from 390.8 to 435.1, a variation percentage of 11.7%, that

was 5% higher than the variation percentage of control group. The second parameter

1 3

h h

, measuring the effect of T wave, demonstrated distinct enhancement of arterial

elasticity in the experimental group (three times the increasing rate of the control

group). We discovered that even though some experimental subjects had vague T

wave before meditation, it was often boosted after meditation. On the other hand, T

control subjects did not even have a T wave before and after relaxation. If T waves

merged with P waves, the parameter

1 3

h h

was considered to be zero. Next, the high

decreasing rate of

1 4

h h

revealed reduced peripheral resistance after meditation. Finally,

the increasing

1 5

h h

infers that Zen meditation significantly improves the quality of

semilunar valves and arterial elasticity. The t-test results showed that all P-values

were smaller than 0.05, further corroborating the significance of the improvement in

the meditation group. In comparison with normal relaxation, Zen meditation may

effectively improve the characteristics of cardiovascular system according to

Table 2-1. The statistical results of four parameters ( 1 1 t h , 1 3 h h , 1 4 h h , 1 5 h h ) and their

variation percentages. P values are evaluated to show the statistical significance

of discrimination between two groups.

1 1 t h 1 3 h h 1 4 h h 1 5 h h Before Relaxation (Mean±Std.) 0 . 26 6 . 380 ± 0.21±0.19 0.27±0.08 0.13±0.04 After Relaxation (Mean±Std.) 3 . 25 7 . 405 ± 0.25±0.19 0.24±0.07 0.15±0.07 Control group Variation percentage % 7 . 6 13.0% −9.3% 14.2% Before Meditation (Mean±Std.) 0 . 28 8 . 390 ± 0.20±0.19 0.25±0.09 0.13±0.04 After Meditation (Mean±Std.) 9 . 18 1 . 435 ± 0.36±0.13 0.22±0.08 0.16±0.03 Experime ntal group Variation percentage % 7 . 11 41.2% −13.5% 27.9% P-value ) 05 . 0 ( 005 . 0 < ( 0.05) 026 . 0 < ( 0.05) 032 . 0 < ( 0.05) 029 . 0 <

Chapter 3

F-VEPs in Zen meditation

Observation of the inner-light perception in deep Zen meditation (Lo et al., 2003)

has aroused our attention. Based on the recording of F-VEPs (flash visual evoked

potentials), this study was thus designed to investigate the characteristics of visual

nervous pathway for the Zen-meditation practitioners (experimental group), in

comparison with that for the normal, healthy subjects (control group). Flash stimuli

were applied before, during and after meditation / relaxation in experimental / control

subjects. We focused on the F-VEPs at the occipital site Oz, central site Cz and frontal

site Fz.

3.1 Background and motivation

During the past decades, a number of papers have reported the benefits of

meditation to the physiological and mental health, with particular emphasis on

transcendental meditation, Yoga meditation, and Japanese-Zen meditation. It is the

first attempt to investigate the electrophysiological signals of the orthodox

Zen-Buddhist practitioners. The main doctrine of Zen-Buddhist practice is to

enables the practitioners to reach a fully egoless, transcendental state (the Alaya state)

(Lo et al., 2003).

Researchers have been probing into the physiological and psychological

parameters during meditation for several decades (Jevning et. al., 1992; West, 1980).

Some important results include: Wallace (Wallace, 1970) claimed the emergence of

theta waves in the frontal area in transcendental meditation, Banquet observed the

slowdown of alpha frequency and the increase of alpha amplitude as well as the

occurrence of the rhythmic theta trains (Banquet, 1973). Recently, Lutz and et al.

found that long-term Buddhist practitioners self-induced sustained high-amplitude,

gamma-band EEG and phase-synchrony during meditation (Lutz and et al., 2004).

These patterns obviously occurred at the lateral fronto-parietal electrodes. Travis

found several physiological markers including the decrease of respiration rate, higher

respiratory sinus arrhythmia amplitudes, higher alpha coherence, etc (Travis, 2001).

In addition, ERPs (evoked response potential) used to explore the underlying

neuron activities in meditation were investigated. Zhang claimed the increase of

F-VEP amplitudes of Qigong practitioners under meditation (Zhang, 1993). In Xu et,

al.’s research, increase of amplitude and decrease of latency were reported (Xu et al.,

1998). Furthermore, auditory evoked potential (AEP) under meditation has also been

(McEvoy et al., 1980; Tells et al., 1994).

The paper survey given above shows a particular phenomenon of meditation, that

is, its effects on the frontal cortex. The transcendental state of Zen meditation, in fact,

reflects that the human life system turns off its physical and mental sensors, leaves off

the message transmission from outside world, and keeps subconscious tranquil. When

further attaining the deeper meditation state, practitioners often ignite their inner

energy, accompanied with the experience of perceiving the inner light (Lo et al.,

2003). Base on the observation of frontal EEG and the common experiences of

inner-light perception, we hypothesized that meditation would affect the visual neuron

pathway, and the effect might be modulated by the meditation effects in the frontal

area of the cortex.

Accordingly, this study aimed to investigate the meditation effects on visual

neuron pathway by quantitatively analyzing the visual evoked potentials (VEPs).

Owing to the limitation that meditators must close their eyes during meditation, we

employed the flashed light as the visual stimulus and recorded the flash visual evoked

potential (F-VEP), with particular emphasis on channels Oz, Cz, and Fz.

3.2 Experimental setup and procedure

subjects (normal, healthy people without any experience in meditation). In the

experimental group, 15 females and 15 males at the mean age of 28.7±4.6 years

participated. Their experiences in Zen-Buddhist practice span 6.6±4.1 years. The

control group consists of 9 females and 21 males at the mean age of 24.1±1.6 years.

EEG and F-VEP were recorded within the frequency range from 0.15Hz to 50Hz. The

sampling rate is 1000Hz. We applied the 30-channel recording montage with the

ground at the forehead and the reference as the linked mastoids.

F-VEPs were recorded before, during and after the main session (meditation or

relaxation), that is called the pre-, mid-, or post-session. Continuous, 100 flash stimuli

were applied to the subject in each session. The flash light was 10 μs in duration and 1 Hz in frequency produced by a xenon lamp that was placed 60 cm in front of the

subjects’ eyes. These parameters were referred to the standard procedures (Adrian and

Mathews, 1934; Cigánek, 1961; Odom et al., 2004). We averaged these 100 trials to

get the averaging F-VEP in each session. Subjects sat in a isolated space during the

recording. Each recording lasted for about one hour. The course included 10min

pre-session, 40min mid-session, and 10min post-session recording. In the mid-session

period, experimental subjects practiced the Zen meditation, while control subjects sat

in normal relaxed position with eyes closed. During the meditation, the subject sat,

the subjects focused on the Navel Chakra and regulated their respiration. Navel

Chakra is regarded as an important switch that activates the other Chakras. After

igniting the Navel Chakra, meditators focused on the Zen Chakra or Dharma Chakra

to empty the thought.

3.3 Results and inter-group comparison

Figures 3-1 and 3-2 are the averaged F-VEPs on Oz, Cz and Fz of one

experimental and one control subject, respectively. The dotted, solid and dashed lines

represent respectively the pre-, mid-, and post-session F-VEPs. Tables 3-1 and 3-2 list

the average ratios of mid- to pre- and post- to mid-session latencies and amplitudes

for both group. The p values are calculated by paired t-test and t-test.

The peak numbers are named according to the visual evoked potential standard

(Odom et al., 2004). From Table 3-1, we can examine the variations in latencies

0 50 100 150 200 250 300 350 400 450 (a) 20 ms -20 uv P2 N3 P3 N2 P1 N4 0 50 100 150 200 250 300 350 400 450 (b) -20 uv 20 ms P1 P2 N3 P3 N1 N2 0 50 100 150 200 250 300 350 400 450 (c) -20 uv 20 ms N2 P1 P2 N3 P3 N1

0 50 100 150 200 250 300 350 400 450 (a) -10 uv 10 ms P2 N3 P3 N4 N2 P1 0 50 100 150 200 250 300 350 400 450 (b) -10 uv 10 ms P2 P3 P1 N1 _N 3 N2 0 50 100 150 200 250 300 350 400 450 (c) -10 uv 10 ms P2 P1 P3 N1 N2 N3

Table 3-1. Average ratios of peak latencies (NS: not significant).

Exp group Ctrl group t-test

Item mid/pre (%) post/mid (%) P value (paired t-test) mid/pre (%) post/mid (%) P value (paired t-test) mid/pre post/mid P1 106.45 98.79 NS 110.79 96.36 NS NS NS N2 100.65 100.16 NS 103.40 99.94 NS NS NS P2 103.52 99.13 0.001* 104.60 100.72 NS NS NS N3 101.91 99.99 NS 100.00 101.59 NS NS NS P3 101.34 99.85 NS 100.24 101.47 NS NS NS Oz N4 100.57 99.81 NS 100.62 99.51 NS NS NS N1 103.47 101.77 NS 129.77 98.57 0.009* 0.008* NS P1 100.72 102.66 NS 109.84 106.50 NS 0.009* NS N2 101.15 99.61 NS 102.68 98.82 NS NS NS P2 101.03 99.93 NS 102.51 100.00 NS NS NS Cz N3 103.41 100.09 NS 105.02 100.40 NS NS NS N1 105.77 104.65 NS 111.05 92.74 0.040* NS 0.042* P1 102.32 100.75 NS 109.13 97.42 0.024* 0.024* NS N2 100.98 100.38 NS 102.19 99.59 NS NS NS P2 100.50 99.30 NS 103.65 99.47 NS NS NS Fz N3 99.81 100.29 NS 105.68 98.30 0.021* 0.042* NS

We summarize some important findings as follows (exp: experimental group, cnt:

control group).

(1) Significant variations in latency are: (exp) P2 at Oz; (cnt) N1 at Cz, P1 and N3 at

Fz. Note that all the latencies of these components increase during

meditation/relaxation and decrease in the post-session.

(2) Two groups show different trends in the latency ratios: (mid/pre) N1 and P1 at Cz;

P1 and N3 at Fz. The latencies of these components in the control group show

considerable increasing, whereas the changes of experimental group are not

apparent. (post/mid) N1 at Fz: the latency of the experimental group increased, but

Table 3-2. Average ratios of peak amplitudes (NS: not significant).

Exp group Ctrl group t-test

Item mid/pre (%) post/mid (%) P value (paired t-test) mid/pre (%) post/mid (%) P value (paired t-test) mid/pre post/mid P1-N2 112.89 102.20 NS 121.64 100.98 NS NS NS N2-P2 104.98 104.86 NS 110.82 115.67 NS NS NS P2-N3 109.57 100.72 NS 121.50 116.99 NS NS NS N3-P3 82.93 128.93 0.016* 124.61 111.83 NS 0.006* NS Oz P3-N4 89.10 126.96 0.037* 132.79 116.69 NS 0.002* NS N1-P1 95.91 114.92 NS 108.88 103.98 NS NS NS P1-N2 112.49 90.94 NS 84.88 117.91 0.038* 0.019* 0.031* N2-P2 102.87 107.42 NS 98.14 115.47 NS NS NS Cz P2-N3 125.30 119.52 NS 113.01 130.20 NS NS NS N1-P1 112.13 113.35 NS 106.25 109.02 NS NS NS P1-N2 100.18 101.09 NS 84.82 113.64 0.047* 0.045* NS N2-P2 99.62 110.85 NS 105.07 103.83 NS NS NS Fz P2-N3 105.58 122.10 NS 121.35 114.16 NS NS NS

From Table 3-2, we can examine the variations in amplitudes between sessions or

between groups. We summarize some important findings as follows.

(1) Significant variations in amplitude are: (exp) N3-P3 and P3-N4 at Oz; (cnt) P1-N2

at Cz and Fz. The amplitudes of these components decrease during

meditation/relaxation and increase afterwards.

(2) Two groups show different trends in the magnitude ratios: (mid/pre) N3-P3 and

P3-N4 at Oz; P1-N2 at Cz and Fz; (post/mid) P1-N2 at Cz. In experimental group,

the amplitudes of N3-P3 and P3-N4 at Oz are decreased under meditation, but the

trends are opposite under relaxation in the control group. The amplitudes of

P1-N2 on Cz and Fz are increased in experimental group but decreased in the

control group during meditation/relaxation. Furthermore, after

meditation/relaxation, P1-N2 at Cz is decreased in the experimental group and is

Chapter 4

Spatial Focalization of

Zen-Meditation Brain Based

on EEG

The aim of this study is to report our preliminary results of investigating the

spatial focalization of Zen-meditation EEG (electroencephalograph) in alpha band

(8-13 Hz). For comparison, the study involved two groups of subjects, practitioners

(experimental group) and non-practitioners (control group).

To extract EEG alpha rhythm, wavelet analysis was applied to multi-channel EEG

signals. Normalized alpha-power vectors were then constructed from spatial

distribution of alpha powers, that were classified by Fuzzy C-means based algorithm

to explore various brain spatial characteristics during meditation (or, at rest). Optimal

number of clusters was determined by correlation coefficients of the

membership-value vectors of each cluster center.

4.1 Background and motivation

meditation. Scientists have corroborated that meditation considerably lowered down

the levels of respiration rate, heart rate, spontaneous skin conductance response and

cortisol (Dillbeck and Orme-Johnson, 1987; Jones, 2001). Other researches reported

that meditation was useful in treating some medical problems such as hypertension

(Walton et al., 1995; Barnes et al., 2001), anxiety (Lindberg, 2005), pressure (Shetty,

2005), and even tumors (Ott et al., 2006). In psycho-neurology, EEG becomes an

important tool to monitor the meditation effects on the neural systems.

According to those literatures, the most common phenomenon is the increase of

alpha-band power and alpha coherence in the frontal areas (Aftanas and Golocheikine,

2001; Travis et al., 2002; Murata et al., 2004; Takahashi et al., 2005). Alpha activities

are normally recorded on the posterior scalp regions for normal, healthy adults with

eye-closed relaxation under conditions of physical relaxation and relative mental

inactivity. In Cantero et al.’s research, they concluded that alpha rhythm could be the

baseline of brain activity when sensory inputs are very few at the state of relaxed

wakefulness (Cantero, Atienza, Salas and Gómez, 1999). Cantero et al. provided

evidence for the alpha power modulation and different scalp distributions at a

particular cerebral arousal state (Cantero, Atienza, Gomez and Salas, 1999).

Correlation between cognition and alpha variation during meditation was studied by

during meditation reflected the increasing power of slow alpha in the frontal areas

(Aftanas et al., 2001). Since alpha variations are the common characteristic during

meditation, investigation of alpha spatial-temporal traits is important for further

understanding the neural-physiological effects of meditation.

Among various meditation techniques, we focus on Zen meditation that has been

becoming popular in Taiwan. During meditation, meditators first transcend their

physical and mental perceptions via particular mind-focusing technique, leave off

the message transmission from outside world, and concentrate their attention on

some particular Chakras (Lo et al., 2003).

Exploration of the meditation alpha activities has been focused on the change of

alpha power and coherence, with few studies on the spatial-temporal behavior. Owing

to the particular cognitive significance of alpha rhythms, investigating the distribution

of alpha band might help understanding the meditation effects on the cortical

activation. The aim of this study is thus to investigate the temporal evolution of

spatial characteristics of alpha rhythms under Zen meditation.

In the following section, we introduce the methods for multi-channel meditation

EEG processing and analysis, including wavelet transform for identifying

alpha-dominated sections and Fuzzy C-means for classifying the feature vectors that

of spatial characteristics between experimental group (Zen meditators) and control

group (non-meditators).

4.2 Proposed schemes

4.2.1 Wavelet Transform

Spectrum analysis based on Fourier Transform (FT) has been the most popular

method for identifying various EEG rhythms. However, FT can not resolve the

time-varying spectral properties of EEG. Wavelet Transform (WT) is one of the

approaches developed to solve the problem (Daubechies, 1992). WT decomposes a

signal into scaled, time-shifted version of the pre-designed wavelet prototype. WT

possesses the capability of local analysis as well as high flexibility in terms of

scalability in resolution. However, choosing an appropriate wavelet prototype is

always the issue firstly encountered. A rational thought is to choose a wavelet model

with its pattern matching the shape of the component to be analyzed. Accordingly, the

wave shape of Daubechies 6 (Db 6) low-pass filter was selected in this study.

Furthermore, Daubechies wavelets are used due to their properties, including good

regularity for high number of moments (Dragotti and Vetterli, 2003).

4.2.2 Alpha detection

frequency domains. Due to the time-varying spectral behavior of alpha patterns, a

wavelet-based algorithm becomes appealing for alpha detection.

To decompose the EEG, discrete wavelet transform (DWT) was implemented by

the typical pyramidal structure (order 2), with a window size of 1 second and no

overlap. We then calculated the wavelet coefficients corresponding to the delta (δ: 3.5 Hz or less), theta (θ: 4~7.5 Hz), alpha (α: 8~13 Hz), beta (β: 14~30 Hz) and gamma (γ: above 30 Hz) band. Finally, EEG rhythm in each particular band was reconstructed. Define the alpha-power percentage ρ below,

%

100

×

+

+

+

+

=

γ β α θ δ α

ρ

p

p

p

p

p

p

(4.1)

where Pα denotes the power of the reconstructed α wave, and so on.

Consider an EEG epoch. If the alpha-power percentage ρ is greater than a

pre-defined threshold θ1 (θ1=50% in this study), it is considered as an

alpha-dominated epoch. Figure 1 displays alpha-power percentages varying with time.

From Figure 1, alpha apparently dominated the last 3 seconds (ρ > 50%), that

were to be extracted for further spatial analysis. All the 30-channels were examined to

identify the alpha-dominated epochs. Those epochs with at least one channel

satisfying ρ>50% were extracted. Next the alpha-power vector is defined as

]

[ _{i1 i2 i3 i30}

i p p p p

p_α = _{α α α} " _α (4.2)

element Pαij representing the alpha power of the jth channel. Final feature vector was obtained by normalizing the alpha-power vector based on the pool of all vectors

extracted.

ρ(%)

sec

Figure 4-1: A section of 5-sec EEG. The numbers listed below are the alpha-power percentages.

4.2.3 Fuzzy C-means

Fuzzy c-means (FCM) is a data clustering method wherein each data point belongs to

a cluster to some degree specified by a membership value (Bezdek, 1973). We utilized

FCM to classify the input feature vectors (normalized alpha-power vectors). In FCM,

number of clusters needs to be determined first. We proposed the idea of evaluating

correlation coefficients to estimate the appropriate number of clusters. First, with an

initial guess (>5), we can derive the membership value for each sample xj:

1 1 1 2 − = −

⎥

⎥

⎥

⎦

⎤

⎢

⎢

⎢

⎣

⎡

⎥

⎥

⎦

⎤

⎢

⎢

⎣

⎡

=

∑

c l lj ij ij

d

d

β

χ

_(4.3)

where χij is the membership value of sample xj with the center yi, dij is the distance

center yi has its membership-value vector,

]

[

_{i1 i2 im}

i

χ

χ

χ

χ

=

"

(4.4)

Here m is the number of extracting vectors. The correlation coefficient is

)

,

(

)

,

(

)

,

(

)

,

(

j j i i j i j i

C

C

C

R

χ

χ

χ

χ

χ

χ

χ

χ

=

_(4.5)

where C(

χ

i,

χ

j)=E

[

(

χ

i−

μ

i

)

(

χ

j−

μ

j

)

]

is the covariance matrix.

If R(χi,χj) is larger than the threshold θ2 (in statistical reason, we set θ2 =0.3 for the

side line between strong and weak correlation), yi and yj are too close, and the number

of cluster c must be reduced by 1.

First we checked the performance of our algorithm by analyzing 800 real EEG

epochs. Normalized alpha-power vector is illustrated by colored brain mapping.

Figure 4-2 displays the results of classifying the alpha-power mappings into 4 clusters.

Figure 4-2 and 4-3 display some sample brain mappings in each cluster for 4 and 3

clusters, respectively. Table 4-1 lists the correlation coefficients between clusters i and

Figure 4-2: The results of 4 clusters

Table 4-1: The correlation coefficients (4 clusters)

Cluster 1 2 3 4 1 0.67 -0.86 -0.32 2 0.67 -0.73 -0.64 3 -0.86 -0.73 0.04 4 -0.32 -0.64 0.04 Cluster #4 Cluster #3 Cluster #2 Cluster #1

Apparently, mappings in cluster #1 are very similar to those in cluster #2, and a

large correlation coefficient R(χ₁,χ₂)=0.67 is obtained. Accordingly, number of clusters should be reduced. Figure 4-3 shows the results of 3 clusters. Table 4-2 shows

the correlation coefficients for the 3-cluster case. Note that none of these coefficients

exceeds θ2, that indicates the number of clusters c = 3 is appropriate for the sample

pool analyzed. The example above demonstrates the strategy of determining the

appropriate number of clusters. Flowchart of the algorithm is shown in Figure 4-4.

Figure 4-3: The results of 3 clusters

Table 4-2: The correlation coefficients (3 clusters)

Cluster 1 2 3 1 -0.80 -0.51 2 -0.80 -0.11 3 -0.51 -0.11 Cluster #3 Cluster #2 Cluster #1

30-Channel EEG data

Alpha wave detection

Classification based on FCM

Examine the cluster number by checking

(

i j

)

R χ ,χ Yes No R

(

χ_i,χ_j

)

< c = c - 1 ∀ i, j θ2

Output the results of classification

4.3 Subjects and recording setup

This study involved 10 experimental subjects (Zen meditators) and 10 control

subjects (normal, healthy people without any experience in meditation). Experimental

group included 3 females and 7 males at the average age of 28.9±3.2 years. Their

experiences in Zen-Buddhist practice span 7.6±4.5 years. Control group consisted of 4

females and 6 males at the average age of 25.9±5.2 year.

EEG was recorded within the frequency range from 0.15Hz to 50Hz, with a

sampling rate of 512 Hz. We applied the 30-channel recording montage, based on the

10-20 system, with the ground at the forehead and the linked mastoids as the

reference.

Subjects sat in an isolated space during the recording. Each recording lasted for

about 34 minutes, including 2-minute pre-session, 30-minute main-session, and

2-minute post-session recording.

In the main-session period, experimental subjects practiced the Zen meditation, while

control subjects sat in normal, relaxed position with eyes closed. During the

meditation, experimental subject sat in the full-lotus or half-lotus position, with eyes

closed. While before and after the main-session period, both experimental and control

4.4 Results

Figs. 4-5 to 4-7 plot the results for one experimental subject (cluster number c=3).

Totally 709 alpha-power vectors were detected and analyzed. Only a portion of

alpha-power brain mappings are displayed for each cluster.

Table 4-3 lists the Euclidean distances between different cluster centers. Standard

deviation of the distances of all vectors away from the cluster center is computed for

each cluster (Table 4-4). According to Tables 4-3 and 4-4, distances between cluster

centers are greater than three times of the standard deviations. Evidently, FCM

successfully separated these clusters.

Table 4-3: The Euclidean distances between cluster centers

Cluster \ Cluster 1 2 3

1 1.474 0.765

2 1.474 0.895

3 0.765 0.895

Table 4-4: The standard deviation of the distances of each sample to its center

Cluster Number of

samples Standard deviation of the distances

1 266 0.250

2 246 0.232

To investigate the spatiotemporal evolution of alpha activities, we employed the

color-chart illustration with red, blue and green indicating the emergence of brain

mappings belonging to cluster #1, #2 and #3, respectively (Figure 4-8). Black

indicates the non-alpha epochs.

From this chart, we could assess the alpha distribution easily. For instance, we

could find frontal alpha (class #1) emerged more often in the middle and non-alpha

presented in the late phase of the main-session. The statistical analysis of

spatiotemporal distribution of alpha activities for both groups will be discussed in the

next section.

Figure 4-5: Selected samples (three rows) and the center of Cluster #1. The numbers above the samples are the time indices (in sec).

Figure 4-6: Selected samples (three rows) and the center of Cluster #2. The numbers above the samples are the time indices (in sec).

Figure 4-7: Selected samples (three rows) and the center of Cluster #3. The numbers above the samples are the time indices (in sec).

Pre-session 0 20 40 60 80 100 120 (sec) 220 420 620 820 1020 1220 1420 1620 120 320 (sec) 320 520 (sec) 520 720 (sec) 720 920 (sec) 920 1120 (sec) 1120 1320 (sec) 1320 1520 (sec) 1520 Main-session 1720 (sec) 1820 1720 1920 (sec) Post-session Not α-dominant 1920 2040 (sec) Cluster #2 Cluster #1 Cluster #3 Figure 4-8: The color chart of alpha distribution