利用腦電磁訊號之功能性連結時空造影

Abstract

為瞭解大腦如何運作，大腦功能性連結方面的研究甚是重要，目前已有證據顯示不正常的功能性連結會導致精神疾病和腦部疾病。由於腦電圖及腦磁圖具有高時間解析度，故特別適合用於功能性連結的研究，但在計算感測器間的功能性連結時，電導體傳導(volume conduction)和磁場延伸(field spread)會造成偽連結的問題，在感測器層級所估計的連結，無法顯示出腦部皮質對皮質或是皮質對次皮質之間的關聯性，因此在活動源階層估算腦部功能性連結是必要的。本研究目的在於建立一套腦部功能性連結時空造影的系統，可準確的定位，具有高感測度及高效率。獨立成分分析法一般用於去除雜訊並提高訊雜比，本論文提出一創新的區域間關聯性估測法，其利用大腦標準分區及獨立成份之皮質活動，以統計的方式計算出區域間的關聯性。腦電磁波訊號源造影的眾多方法中，以光束造影法最為有效且正確，本論文發展出令光束造影法可用於偵測關聯性活動源和相位振幅耦合活動源的技術。線性相關活動源成分時空造影法(SILSC)與最大化複相關活動源光束構成法(MMCB)可估算參考訊號與腦內各個活動源間的時序相關程度，以光束構成法為基礎之相位振幅耦合造影法(BIPAC)則可計算指定的參考訊號與活動源間的相位振幅耦合程度，此三種光束構成法皆提供了活動源方向的解析解(analytical solution)，相較傳統方法可更準確。本研究利用模擬資料來進行以上四種方法的效能準確度評估，並利用性別辨認及臉部表情模仿實驗的腦磁波資料來說明演算法的應用性。在模擬實驗中，四種造影方法皆可準確地定位活動源，三種光束構成法對與參考訊號相關的活動源有高敏感度，且相較於傳統方法活動源的空間分布更集中，空間散佈範圍較小。在性別辨認實驗中，以區域間關聯性估測法來建立功能性網路，結果呈現出臉部辨認核心系統內的各腦區之間有高度連結，說明區域間關聯性估測法可找出與任務相關的腦區彼此間之活動關聯。在臉部表情模仿實驗中，我們以額下回(inferior frontal gyrus)做為參考，利用線性相關活動源成分時空造影法來估算各腦區與額下回的動態關聯，在模擬表情時，呈現高關聯值的區域為鏡像神經元系統相關區域，當執行產生表情時，呈現高關聯的腦區則是運動控制相關的區域，該差異顯示了額下回的功能分化，因此線性相關活動源成分時空造影法亦可作為研究功能分化的工具。本論文提出的四種功能性連結造影法，提供了以不同評量方法所計算的腦區間功能性連結，可用於研究以不同機制進行的訊息傳遞，未來亦期望用於研究精神疾病和腦部疾病的病理，進一步找出生物標記以輔助醫師進行診斷及病程追蹤。

Investigation of functional brain connectivity is crucial to understand how the brain works. There is emerging evidence that abnormal functional connectivity can cause metal diseases and brain disorders. Magnetoencephalography (MEG) and electroencephalography (EEG) are particularly suitable to examine connectivity because of their high temporal resolution. Computing the functional connectivity at MEG/EEG sensor level is exposed to the problem of spurious connectivity caused by volume conduction and field spread. Moreover, sensor-to-sensor connectivity cannot reveal cortico-cortical or cortico-subcortical interactions. Therefore, assessing the functional connectivity at source level is essential. This research aims at developing a framework for spatiotemporal imaging of functional connectivity with high localization accuracy, sensitivity, and efficiency. Independent component analysis (ICA) is widely used to remove artifacts and increase signal-to-noise ratio. In this thesis, we proposed an inter-regional association estimation method, which utilizes the cortical mapping of independent components and the brain atlas to statistically quantify the relationships between structural regions. For imaging the source activity from MEG and EEG, beamforming methods are demonstrated to be effective and accurate. We proposed three techniques to improve the performance of beamforming methods with respect to source localization as well as extend the functionality of beamformer to the imaging of correlation and phase-amplitude coupling (PAC). Spatiotemporal imaging of linearly-related source component (SILSC) and maximum multiple-correlation beamformer (MMCB) can quantify the interdependencies according to correlation between estimated source component and the specified reference signals. Beamformer-based imaging of phase-amplitude coupling (BIPAC) can quantify the phase-amplitude coupling between the reference signal and the estimated source component. The three beamformer-based methods can compute the source activity with an analytical solution of dipole orientation. Accuracy performance of these proposed techniques was examined by simulation studies. The applicability and feasibility of inter-regional association estimation method and SILSC were shown in the gender discrimination and facial expression imitation experiments, respectively. In the simulation studies, the four proposed imaging methods can accurately localize sources. Moreover, the three beamformer-based imaging methods can achieve less spatial spread than conventional methods and have high sensitivity to the sources having activity related to the reference signal. In the gender discrimination study, inter-regional association estimation method was applied to estimate the functional network and the regions revealing strong association values have been reported to be the core system of face perception. The result demonstrates that the proposed inter-regional association estimation method can detect the task-related regions. In the facial expression imitation experiment, SILSC was applied to estimate the dynamics of correlation between the inferior frontal gyrus (IFG) and other regions. When imitating and executing facial expressions, the regions related to mirror neuron system and motor control revealed high level of correlation to the IFG, respectively. The difference between two conditions demonstrates the functional segregation of IFG and SILSC can be the tool for investigating functional segregation. The four proposed imaging methods assess the functional connectivity using different measures and can be used to investigate different mechanisms of inter-regional communications. These methods are expected to investigate the pathology of mental illness and brain diseases as well as detect the biomarkers to benefit the diagnosis and progress tracking of these diseases.