禪定以及休息之Alpha波和腦電位質心的微狀態研究

Abstract

本論文主要提出一套分析的方法來探討禪定與一般放鬆休息之腦電波的微狀態特性的差異性，特別針對空間微狀態特性隨著時間的演化。在這研究中發展了兩種不同的方法以探討微狀態。第一個方案CWT-NSAD，是基於α能量的空間分佈。首先腦電波透過連續小波轉換(Continuous wavelet transform)分解成5個頻帶(∆、θ、α、β、γ)。α能量的百分比被用來判斷是否為α-dominant。對於所有的α-dominant的α能量(歸一化)組成的特徵向量經由NSAD(normalized sum of absolute difference)進行分類。本論文中提出的策略成功地將α能量(歸一化)的腦電波地形圖進行分類，分類的結果優於FCM(Fuzzy C-means)。最後分類結果可被應用於解讀與分析長期腦電波的微觀行為。 第二個方案為PMC (mass center of peak)，此分法是直接計算頻道Fz主要峰值的電位場質心。頻道Fz經常受到眼球運動的影響而造成嚴重的基準線飄移。我們利用線性回歸(linear regression)以進行腦電波訊號的基準線校正。經過基準線校正後，峰值檢測被用來識別大量主要的正(負)峰值，PMC的微觀分析是針對主要的峰值進行。每個瞬時微觀狀態是針對峰值質心的計算經由空間幾何座標量化。最後我們可以從大腦中的五個區域(frontal、left temporal、right temporal、central以及posterior)來探索微狀態的轉變現象。 我們的初步研究結果，CWT-NSAD揭露了禪定腦電波在整個禪定過程中α能量水平顯得格外一致，特別是在前額葉區域。在PMC的部份，禪定腦電波相對於休息腦電波(一分鐘內46次轉換)顯示了在frontal以及central快速和頻繁的微狀態轉換(一分鐘內68次 FC 或者CF)。此外，禪定組的Positive PMC發生在posterior區域的次數(機率:2.3%)少於控制組(5.6%)。

This thesis is aimed to investigate the temporal evolution of spatial microstates of 30-channel Chan-Ding and resting EEG (electroencephalograph). Two different schemes of microstate analysis are developed in this study. The first scheme, CWT-NSAD, is based on spatial distribution of alpha power. The EEG signal is first decomposed into five EEG rhythms (∆, θ, α, β, and γ) by continuous Wavelet Transform (CWT). The percentage of  power is used to identify whether a 0.25-second epoch is α-dominant. For all the -dominant EEG epochs, the feature vectors composed of 30 normalized-to-unity (NU) alpha powers are classified by NSAD (normalized sum of absolute difference) based strategy. The strategy proposed in this thesis successfully classifies the EEG brain mappings of NU alpha powers. Performance of NSAD-based strategy is superior to FCM (fuzzy C-means) clustering. Finally, the classification results can be employed in long-term EEG interpretation and analysis of microstate behaviors. The second scheme, PMC (mass center of peak), directly evaluates the mass center of the brain electrical-potential field constructed for the major peaks of channel Fz. Channel Fz often picks up eye-motion artefacts that cause serious baseline drift. We apply linear regression to EEG baseline-drift correction. After the baseline correction, peak detection is employed to identify the major positive and negative peaks of substantial amplitude. The PMC microstate analysis is conducted on the major peaks. Each instantaneous microstate is quantified by the spatially geometric coordinates of centers of mass evaluated for major positive and negative peaks (positive and negative PMCs). Finally, we may explore the microstate behaviors from the spatial transition of positive and negative PMC in five regions (frontal, left temporal, right temporal, central, and posterior region). In our preliminary results, CWT-NSAD reveals the extraordinarily consistent -power level through the entire Chan-Ding EEG record, particularly in the frontal region. PMC discloses the rapid and frequent microstate transitions between frontal and central regions (68 FC or CF in one minute) in Chan-Ding EEG compared with the resting EEG (46 transitions in one minute). In addition, Chan-Ding positive PMC visits the posterior region (probability of 2.3%) less often than resting PMC (5.3%).