| 摘要: | 近年來，以經驗模態分解為基礎的方法，例如經驗模態分解法( Empirical Mode Decomposition)、總體經驗模態分解法(Ensemble Empirical Mode Decomposition )、多變量經驗模態分解法(Multivariate Empirical Mode Decomposition)與互補總體經驗模態分解法(Complementary Ensemble Empirical Mode Decomposition)等常被應用在萃取醫學應用之非穩態的信號，如分析血壓、心電圖心跳速率變化、肺動脈高血壓、腦部波介面以及功能性磁振造影的血氧濃度相依信號等。經驗模態分解法可將信號分解成有限的本質模態函數(intrinsic mode functions, IMF)，以往研究顯示，經驗模態分解法是一種資料驅動的方法，並適用於萃取隨機訊號。但經驗模態分解法對於突然變化或間斷的信號會有模式混合的現象，導致萃取本質模態函數出現異常。然而總體經驗模態分解法(EEMD)處理訊號時，須經由大量重複測試雜訊添加的信號移位過程，以往的研究經驗，使用總體經驗模態分解法去除殘餘雜訊信號的過程非常耗時。
在此篇論文中，首先我們開發以互補總體經驗模態分解法(Complementary Ensemble Empirical Mode Decomposition)，分析多頻道腦電磁儀(MEG)的信號，萃取受試者聽覺穩態誘發磁場並分解成本質模態(IMFs)，經由與空間模板(Spatial Template)比對，匹配出與聽覺穩態磁場高度相關之本質模態，最後重組成去除雜訊後的穩態聽覺磁場。另外，實驗的第二部分，使用多變量模態分解法(Multivariate Empirical Mode Decomposition)，可將功能性磁振造影每一張影像上的血氧濃度水平依賴信號(Blood Oxygen Level Depend Signals) 分解成共同特徵的本質模態函數，計算本質模態與原始血液熱動力學(Hemodynamic response)之間的相關係數(correlation coefficient)，匹配出與嗅覺刺激血液熱動力學高度相關之本質模態函數，主要目地是重組這些被匹配出的本質模態函數，獲得去除雜訊或人工假影的影像。
;In recent years, Empirical mode decomposition (EMD)-based methods, i.e., EMD, ensemble empirical mode decomposition (EEMD), complementary ensemble empirical mode decomposition (CEEMD) and multivariate empirical mode decomposition(MEMD) have been used to extract nonstationary signals in many applications, such as analysis of blood pressure, detection of heart-rate variability in electrocardiogram (ECG) , pulmonary hypertension, brain computer interface, BOLD fMRI signals , and etc . The EMD approach decomposes a signal into a finite number of intrinsic mode functions (IMF) by iteratively conducting the sifting process, which has been demonstrated as a powerful data-driven tool for extracting meaningful stochastic signals. But EMD is also very sensitive to any unexpected changes in the signal like in case of missing signal components in certain time intervals, and can lead to mode mixing which impeded the interpretation of the extracted IMFS. The EEMD approached which repeatedly performs the shifting process on a noise-added signal for a mass of trials. However, the reduction of residual noise in EEMD is time-consuming which requires a large amount of trials for average.
In this dissertation, firstly, we developed a complementary ensemble empirical mode decomposition (CEEMD)–based approach to extract steady-state auditory evoked fields (SSAEF) in multi-channel MEG data. The CEEMD utilizes noise assisted data analysis (NADA) approach by adding positive and negative noise to decompose MEG signals into intrinsic mode functions (IMF)，By correlating each IMFs with the multi-channel MEG data, the spatial distribution of each IMF can be obtained. Pertinent SSAEF-related IMFs were then chosen through a template matching process to reconstruct noise-suppressed SSAEFs. Secondly, we adopted multivariate empirical mode decomposition (MEMD) to extract olfactory-related features in fMRI BOLD signals. The MEMD enables common features of different scales in an image slice to be arranged in distinct IMFs, so that the task-related signals can be selected and reconstructed. And the noise and artifacts can be removed from reconstructed BOLD signals by deselecting task-unrelated components. |
