多通道禪定與放鬆休息腦電波之相位結構

Abstract

雖然現代醫學及量測儀器日益進步，現今科學對人類大腦的了解仍僅能窺探一二，腦電波仍然難以被完整詮釋並連結到人類的生理、心理與意識狀態，許多研究嘗試開發不同演算法來探討腦電圖中的時域、空間頻譜特性，腦電波中的相位關係卻鮮少被討論。然而，腦電波之空間相位可以反應出大腦的活動位置，因此本研究將分析多通道腦電波之相位結構。 本研究測量30個通道的腦電波訊號，並比較禪定腦電波（實驗組）和休息放鬆腦電波（對照組）的相位結構。首先，我們利用希爾伯特變換(Hilbert-Huang Transform, HHT)計算每一個通道的瞬時相位，在任何時刻中皆可從30個通道中計算出30個相位。由於雙側對稱性和前後關係是腦電波研究的重要性質，因此我們計算橫向（橫向電極陣列）和縱向（縱向電極陣列）腦電波相位差之梯度，得到橫向相位差曲線和縱向相位差曲線，並觀察實驗組與對照組之曲線差異。為了進一步量化相位差資料，我們使用線性回歸方法比較相位差資料之斜率與相關係數。最後，我們使用連續小波轉換(Continuous wavelet transform, CWT)將原始訊號中的Alpha波頻率分離，採用與上述相同的方法獲得縱向和橫向的相位差曲線。比較禪定腦電波和休息放鬆腦電波的相位差曲線，並將其斜率做成直方圖，我們發現禪定曲線的大部分相位差曲線的斜率接近於0。 關鍵字：腦電波、禪定、相位差

Despite the rapid progress of medical technology and measuring instruments, electroencephalograph (EEG) signal is still too complicated to be interpreted and correlated to various physiological, mental, psychological and conscious states. Hence, a number of researches are devoted to the development of different algorithms for exploring the spatio-temporal-spectral features embedded in EEG. Up to now, the phase property is seldom studied. Nevertheless, the spatial phase relation among EEG recording sites may reflect the source focalization of brain activity. This thesis presents our results of analyzing the phase structure of multi-channel electroencephalogram. This study analyzed the phase structures of 30-channel Zen-meditation EEG (experimental group) and resting EEG (control group) and compared the results. Firstly, instantaneous phase of each EEG channel was estimated by discrete Hilbert transform (DHT). For any given time instant, thirty phases were computed for 30 EEG channels. Since the bilateral symmetry and frontal-posterior relation are important features in EEG study, we then calculated the gradient phase differences between EEG channels located along the transversal (trans electrode array) and longitudinal (chain electrode array) directions. As a consequence, we obtained a bunch of curves named as trans phase-difference (PD) curves and chain PD curves. We observed the linear behavior of PD curves. To further quantify the linear feature, line fitting based on linear regression was applied to PD curves. The slope and resulted R value are compared. Finally, the phase structure property of alpha component was studied. Alpha rhythm was extracted by CWT- (Continuous wavelet transform) based bandpass filter. The same scheme as described above was adopted to obtain the trans and chain PD curves. Histograms of the slopes of alpha PD curves were compared between Zen-meditation EEG and resting EEG. The results reveal that most alpha PD curves of Zen-meditation curves have the slope close to 0