利用導場向量空間投影進行腦部獨立活動訊號源之時空造影

Abstract

腦磁圖儀(Magnetoencephalography)及腦電圖儀(Electroencephalography) 利用非侵入式感測器量測因腦部神經元活化所產生的磁場及電位差。為了解腦部協調及控制行為的機制一般均以逆估算演算法(inverse algorithm) 分析感測器所量測的訊號以得知大腦皮質層的活動動態。逆估算演算法可建立於導場向量空間(lead field vector space)模型的假設上且利用條件限制以求得大腦皮質層活動動態分佈。為更精準求得大腦皮質層活動動態分佈，必須去除人體內在或外在雜訊對腦部活動訊號的干擾。若假設腦部活動訊號與雜訊互相獨立，則獨立成分分析(Independent Component Analysis)可運用於分離腦部活動訊號及其干擾源。此外，獨立成分分析亦可求得獨立訊號對應之感測器空間(sensor space)活動動態。此論文利用奇異值分解(Singular Value Decomposition)導場矩陣(lead field matrix)得到一組最足以代表在感測器空間中獨立時序訊號之動態分佈(mixing matrix)的基底(basis)並運用獨立成分分析產生在感測器空間中獨立時序訊號之動態分佈以求得大腦皮質層活動動態。此研究突破獨立成分分析演算法不具有時空造影之限制，且模擬實驗驗證此方法可精確得到獨立活動訊號在大腦皮質層活動動態分佈。

Magnetoencephalography (MEG) and Electroencephalography (EEG) are the non-invasive instruments that record the induced magnetic field and scalp electrical potential. To study the functionality of human brain, inverse algorithms, involves forward model in lead field vector space, are commonly used for estimating cortical source distribution. For more precisely estimation, interferes, such as artifacts and environmental noises, must be removed. Independent Component Analysis (ICA) can be used to remove interferes which are assumed to be independent to brain acitivities. Moreover, ICA also provides the scalp topography, or said the mixing matrix, of components. The proposed method aim to find the cortical source distribution of given independent components by find a set of basis that best represents the mixing matrix using Singular Value Decomposition (SVD). It provides the spatiotemporal imaging of independent brain activities that cannot obtain from ICA. It is demonstrated that the proposed method can provide accurate cortical source distribution from experiment results of simulations.