| 摘要: | 運動想像(MI)是腦機介面使用的腦波特徵之一，常與動作執行(ME)及動作觀察(MO)一起討論，因為都會誘發運動皮質區的Mu節律事件相關失同步(ERD)。過去研究對於這三個運動任務是否有相同源定位有很多的討論，因此本研究的第一個目的是對三個運動任務做源識別。此外，近年來的EEG採集設備往方便性較高的乾電極發展。為利於腦機介面的應用及推動，本研究的第二個目的是利用運動的獨立成分特徵比較乾濕電極的收集訊號能力。
實驗一採用包含ME、MI跟MO的受試者內實驗設計，蒐集13位受試者的64通道濕電極腦電圖，利用兩種不同順序的獨立成分分析(ICA)來對不同運動任務進行源識別。第一，假設不同運動任務有相同的源，聯合三種運動任務的EEG數據做聯合分解，為SD選項；第二，假設三個運動任務有各自的源，分開三種運動任務的EEG數據執行個別分解，分別為ME、MI和MO選項。根據不同分解方式再依據MuERD挑選出不同選項之Mu成分，進一步使用EEGLAB中的DIPFIT2工具箱的四球殼模型做源定位。結果顯示SD選項相較於MO選項在腦中的位置更靠外側；MI選項較ME選項在腦中的位置更靠前側。除此之外，亦發現在不同選項間達統計差異的源定位結果中，頻率峰值也存在不同。因此推論ME、MI及MO應有各自的主導腦區，及符合先前的第二個假設。接著利用實驗一已知的Mu成分來比較乾電極的收集數據的穩定性，實驗二施行了與實驗一相似的實驗設計並收集16位受試者的30通道乾電極腦電圖。預處理乾電極數據時，根據數據特性更改了預處理步驟，並使用聯合分解得到Mu成分。為了避免不同維度影響ICA分解結果，降低濕電極通道數來比較乾濕電極的Mu節律特徵。結果顯示在不同通道的Mu成分辨識結果，並無明顯差異，說明在腦電圖的訊號採集中，不同通道數量對訊號品質的影響較小。此外，乾電極與濕電極在Mu成分辨識無顯著差異，但其中Mu成分的抑制能量及延遲時間有達統計顯著。結果顯示乾電極有能力辨識Mu成分的抑制特性，但信號品質有限不能做源定位等精確研究，而乾濕電極的訊號品質主要差異來自阻抗值的大小，因此降低阻抗值是未來發展乾電極的首要目標。
;Motor imagery (MI) is one of the brain wave features used in the brain-computer interface(BCI). It is often discussed together with the motor execution (ME) and motor observation (MO), because these motor conditions all induce the event-related desychronization(ERD) of the mu rhythm in the motor-related cortex. The relationship between the source locations of where the mu-rhythm of the three motor conditions have been originated has been largely discussed in the literature. Therefore, the first purpose of this study was to identify and compare the sources of the three motor tasks. In addition, EEG acquisition equipment has developed toward more convenient dry electrodes in recent years. To facilitate the application and development of the BCI, the second purpose of this study was to compare the ability of the wet and dry electrodes to collect signals using the mu component characteristics.
First experiment collected 13 participants with wet-electrode EEG. The within subjects experimental design included motor execution, motor imagery and motor observation. Source identification of different motor tasks was performed using two different sequences of independent component analysis. First, the ICA was applied to the preprocessed EEG data with the data from the three different motor condition altogether for the single ICA decomposition(thus, SD); second, the ICA was separately applied to the preprocessed EEG data of each motor condition to obtain ME, MI and MO with ICA being applied to the data of each motor condition separately. Then, the motor-related independent components, maximally projecting to surrounding the C4 channel location with the maximal mu-suppression feature, of each individual subject would be selected and subjected to source localization process using the DIPFIT2 extension under the EEGLAB. If the sources of the three motor conditions were co-located, there should not be any difference in the sources identified with different ICA steps. The results show that the SD ICA is more lateral than the MO ICA, and the MI ICA is more anterior than the ME ICA. Therefore, the mu ERD activity seemed to be originated from different brain regions for the motor execution, motor imagery, and motor observation conditions. In addition, to compare the mu rhythm characteristics obtained using the dry and wet electrodes, the dry electrode EEG of 16 participants were further collected with the performance of the three motor conditions combined. To compare the result of the dry EEG system to that of the wet EEG system, the number of wet EEG channels was reduced from the original 64 channels to 30 channels to match the setting of the dry EEG system. The results indicated that the number of channels had less influence on signal quality. However, the higher impedance inherently associated with the dry EEG electrodes might play more an important role in affecting the signal quality of the acquired EEG signals. Nonetheless, the dry EEG electrodes could still pick up some of the mu-suppression characteristics associated with the MI and other motor conditions. |
