腦電波有效性連結之評估

Abstract

有效性連結為一種探討大腦神經元因果關係的方法，其主要概念是在不同腦區的神經訊號上找尋訊號間相依程度的變化，進而推導腦區之間因果關係。過去最被廣泛適用來分析有效性連結的方法為格蘭傑因果分析（Granger causality），但因為格蘭傑因果分析統計上的特性以及基本假設導致格蘭傑因果分析在研究使用上會有所拘限。轉置熵(transfer entropy)為另一種基於訊息理論用來探討有效性連結的方式，其不須事先假設交互作用模型和非線性的特性，以及格蘭傑因果分析經由數學推導為轉置熵的一種特例，所以轉置熵在神經科學分析上會較格蘭傑因果分析合適。此外，本研究則進一步提出事件相關轉置熵的分析方法，探討因為事件發生所造成的因果關係變化。在實驗部分，本研究提出模擬實驗和維持注意力駕車實驗探討格蘭傑因果分析和轉置熵進行Tonic和Phasic的有效性連結分析的比較。在模擬資料研究結果顯示轉置熵在線性和非線性的因果關係下皆能表現出良好的偵測能力。而在維持注意力駕車實驗中主要是探討開車嗜睡程度和有效性連結之間的關係。Tonic研究結果顯示在前額葉、中央和頂葉區域的因果強度會在任務成效介於中等程度時會有最大的耦合強度，此證據顯示人類為了維持一定程度的任務成效，腦區間的有效性連結會因此增強。在枕部相關連結證據顯示視覺感覺區的連結強度會隨著嗜睡程度增加而減弱。Phasic研究結果顯示，動態刺激的發生在人類最為清醒的時候會產生較強的連結強度，且隨著嗜睡程度的增加此現象會越來越弱；而隨著真實視覺意識到車輛偏移事件，在枕部部分為果連結時會有連結強度的提升；而在做出反應後在大部分的連結都會發生有效性連結的抑制。此研究顯示出轉置熵在分析有效性分析上會能得到較適合的結果，且轉置熵在探討腦和行為關係的神經科學研究上比起傳統腦波分析更進一步深入了解大腦動態連結。

Effective connectivity, which is commonly used in the study of causal relationships, refers to relationships or dependences among neuronal activities in distant brain regions. Granger causality (GC) is a property that is frequently used to estimate effective connectivity. However, GC is limited by inherent assumptions that may lead to incorrect results. Transfer entropy (TE), which is based on information theory, is an alternative metric of effective connectivity. TE is nonlinear and not dependent on a specific model of interaction. GC is a special case of TE for Gaussian variables. TE is more appropriate to use than GC in the field of cognitive neuroscience. This dissertation proposes event-related transfer entropy, which is based on the sliding window approach and transfer entropy, and constructs a measure of dynamic causality for human behavioral responses to stimuli. Simulation experiments were used to demonstrate the effectiveness of TE and GC. A sustained-attention driving experiment was performed to demonstrate changes in effective connectivity of tonic and phasic analysis for drowsy driving. The simulation experiments revealed that TE is superior to GC at detecting both linear and nonlinear causality. Therefore, TE was used to study the effective connectivity in the sustained-attention driving experiment. In the sustained-attention driving experiment, the tonic results indicated that couplings between pairs of frontal, motor, and parietal areas increased at the intermediate level of task performance. This finding suggests that enhancement of cortico-cortical interaction is necessary to maintain task performance and prevent behavioral lapses. The magnitude of occipital-related connectivity monotonically decreased as the performance declined. This result suggests cortical gating of sensory stimuli during drowsiness. The phasic results revealed an increase in the strength of connectivity following the onset of a stimulus (kinesthetic stimulus) when the task performance was high. This finding indicates that enhanced cortico-cortical interaction in response to a kinesthetic stimulus is required for kinesthetic feedback. An increase in the connectivity of the occipital area suggests that human become conscious of visual stimuli suddenly. Most connectivity is suppressed during human behavioral responses. Neurophysiological evidence of brain-behavior relations provide further insight into distributed brain dynamics and the characteristics of brain-behavior relations in operational environments.