| 摘要: | 事件相關電位P300在認知神經科學上廣泛的被應用，它最常利用新異刺激程序所誘發，儘管現有許多研究於探討P300在大腦中產生的機制，特別在它的神經網路與功能上，但到目前為止仍尚未有定論。本研究的目的在於應用事件相關電位之動態因果模型探討體感覺P300在大腦中的神經網路。此外，本研究還利用網路的連結強度與因果關係說明其功能性，當中體感覺P300是以3D虛擬實境所建立的新異刺激程序所誘發。本研究收錄10位健康的右撇子受試者，使他們在虛擬實境下使用慣用手進行接球遊戲，並同時測量腦電波。接球遊戲中有兩種刺激條件，分為標準刺激(有力回饋)與靶刺激(無力回饋)，發生次數分別為479次與121次(機率80%與20%)，藉此誘發體感覺P300。本研究收集30頻道腦電波與2頻道眼動訊號，取樣頻率為250Hz。腦電波離線的處理步驟包含，以刺激發生當下-500至1000毫秒為分段、30Hz的低通濾波器、全自動校正法去除眼動之干擾、最後將相同試驗取平均，以進行傳統的事件相關電位研究。在動態因果模型中，我們取腦電波的0至900毫秒進行主成分分析法，在盡可能保留最大資訊下將維度降低至9，並以此資料進行分析。起初，我們以3篇先前的文獻為基礎，將每篇文獻做單邊性輸入與雙邊性輸入的區分，建構出六種不同的模型，試圖找出最合適的神經網路結構。最後，將最合適的神經網路結構進行Forward、Backward與Forward-backward連結調變的分析，以找出靶刺激與標準刺激的連結差異處。傳統事件相關電位的結果顯示，在虛擬實境下進行接球遊戲確實可誘發體感覺P300。動態因果模型的分析結果顯示，先前文獻所提出的頂葉-額葉網路較能解釋大腦在體感覺新異刺激程序下的活動。在我們的實驗設計下，P300藉由神經網路的Forward與Backward連結調變所產生，當中Forward連結調變的重要性又高於Backward連結調變，Forward的連結調變意味著感覺的訊息、目標的自動偵測、刺激驅使的注意力處理和記憶的更新與儲存所產生的變化。相反的，Backward的連結調變意味著注意力與運動的控制所產生的變化，在我們的實驗設計下，受試者不需對刺激做任何的反應(例如：按按鈕)，因此導致Backward的連結調變是較不重要的。P300, an event-related potential evoked by oddball paradigm, has drawn a lot of attentions in cognitive neuroscience. Many efforts have been put on the study of the generating mechanism underlying P300, in particular, the network architectures and their possible functional roles, yet the conclusion has not been reached.In this study, we aim to explore the hierarchical network of sensory P300 production by dynamic causal modelling for event-related potential (DCM for ERP) with a VR based novel oddball paradigm. Moreover, we investigate the connection strength changes and causal relationship of this P300 network to address the possible functional roles. Ten healthy right-handed volunteers underwent electroencephalogram recording while performing a catch-ball game using their dominant hand in 3-D Virtual Reality environment. For eliciting sensory P300, the game was designed to comprise two types of tactile stimuli: standard (with force feedback via a haptic feedback system) and target (without force feedback), with the 479 and 121 trials (i.e. 80% and 20%occurring), respectively. 30 channels electroencephalogram (EEG) and 2 channels electrooculography (EOG) were recorded with 250 Hz sampling rate during the task. The data were epoched offline, with a peristimulus window of -500 to1000 ms, filtered with 30 Hz low-pass filter, artefact removal using the fully automated correction method and averaged across artefact-free trials for classical ERP study. In DCM analysis, the time window of interest was set from 0 to 900ms, and the data were reduced to nine key dimensions using principal component analysis for computational expense. We first specify six plausible models, differed in the areas and the input based on three previous literatures, to identify the most likely model hierarchy. Then the target-specific modulation effects in terms of forward, backward and forward-backward were tested based on the winning model. Classical event-related potentials analysis suggested that the catch-ball game based on 3-D Virtual Reality can be used to elicit the somatosensory P300 components reliably. DCM results show that the parietal-frontal network was the most possible model among models tested to explain the brain activity during this sensory oddball paradigm. In our experiment, the P300 was generated by the network with forward and backward modulations, and the contribution of forward modulations to the generation of P300 was significantly greater than that of backward modulations. The functional role of this forward modulation may involve in the delivery of the sensory information and the automatic detection of difference, the stimulus-driven attentional processes and the memory operation related to context-updating and subsequent memory storage. In contrast, the backward modulation was thought to engage in attention control and motor control, which is absent (i.e. no motor output requested) in our experiment, and leads to a minor contribution to our model. |
