執行持續性注意力的駕駛任務中動覺回饋與腦波變化之關聯

Abstract

感覺回饋（sensory feedback）是維持注意力（sustained-attention）的重要一環，理論與實驗的證據支持多元（multimodal）感覺回饋能有效減少處於負荷過重的駕駛者之認知負擔（workload）、重拾處於分心狀態的駕駛者之注意力，以及強化駕駛行為上的表現等等。實務上，進階駕駛員輔助系統（advanced driver assistance system）施加以不同形式的感覺回饋來達成警示駕駛者，然而憂慮的是此機制所減少認知需求是否會增加駕駛者心思神遊（mind-wandering）的可能，進而危害安全。本研究欲以腦波（EEG）變化提出神經生理證據，予以佐證在複雜任務與擬真場景中感覺刺激與行為表現之間相互關係。事件相關頻譜擾動（event-related spectral perturbation）顯示，相較於僅視覺輸入（visual input），加入體感回饋（kinesthetic feedback）使EEG頻譜變化與行為失常呈現更高的相關性；動作準備與錯誤偵測期，前扣帶迴（anterior cingulate cortex, ACC）與中扣帶迴（mid-cingulate cortex, MCC）theta（3-7Hz）能量的需求顯著地減少。格蘭傑因果分析（Granger causality）所建立的有效性連結（effective connectivity）顯示，體感回饋使大腦原本執行駕駛任務時所表現出的任務正網路（task-positive network），轉變為預設模式網路（default mode network），此網路好發於心思神遊狀態。以反應時間所衡量的行為表現亦發現，加入體感回饋增加不佳表現試驗與反應時間。最後，根據前述兩研究在神經科學上之發現，爲監控駕駛者的警覺狀態，本研究提出一新穎的分類演算法，特點是融入多重腦區之腦波變化以促成腦機介面（brain-computer interface）在駕駛環境中得以應用。

Theoretical and experimental evidence suggests that multimodal sensory feedback effectively reduces the workload of an overloaded driver, captures the attention of a distracted driver, as well as enhances behavioral performance. Commercial driver assistance systems utilize warning signals presented in different sensory modalities to alert drivers to the danger. However, there are valid concerns that over-relying on multimodal sensory inputs may result in mind-wandering. To test this prediction, this study recorded EEG signals from sixteen participants as they performed a sustained-attention driving task. Participants completed two driving sessions (w/ visual vs. w/ visual and kinesthetic input) on separate days. Results indicated that EEG spectral dynamics highly correlated with performance lapses when driving involved kinesthetic feedback. In contrast to the visual input session, kinesthetic feedback reduced theta-power augmentation in the midcingulate cortex (MCC) and anterior cingulate cortex (ACC) for action preparation and error monitoring. The effective connectivity, measured by Granger causality, revealed that the dominant causal hub shifted to from MCC to PCC, reflecting the deactivation of the task-positive network and the activation of the resting-state default mode network. Additionally, decrease in attentional demand, facilitated by kinesthetic feedback, eventually significantly increased the reaction time in the suboptimal-performance state. Finally, based on these neuroscience discovers of pre-stimuli EEG activities, this study proposed a novel classification algorithm for monitoring driver vigilance that characterizes the EEG dynamics of multiple brain areas favored the application of brain-computer interface in driving.