AACS：EEG-fMRI全自動腦電訊號偽跡校正系統

AACS : Development of an Automatic EEG Artifact Correction System for EEG-fMRI

林怡彣(Yi-Wen Lin)

指導教授 ： 陳志宏

國立臺灣大學/電機資訊學院/生醫電子與資訊學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202200226

摘要

近年來，多模態成像技術常被應用在大腦功能研究中，尤其是神經血管耦合相關的研究領域，經常整合不同影像的結果並探討其機制，而同時取得神經與血管訊號對此耦合現象之研究是至關重要的。非侵入性的功能性磁振造影 (functional Magnetic Resonance Imaging, fMRI) 與腦電圖 (electroencephalography, EEG) 之同步掃描技術，雖然可同時紀錄神經電訊號和來自神經活動所誘發之血管訊號，但卻會對腦電圖產生嚴重干擾：(1) 梯度偽跡與 (2) 心衝擊波偽跡。因此，如何將干擾訊號去除是該同步技術的核心問題之一，也因其訊號處理之技術門檻，使得EEG-fMRI同步技術尚未被廣泛應用。

迄今為止，針對腦電圖之偽跡雖然已提出多種校正方式，但對於何種方法能在校正後得到最佳數據質量，仍未有共識。而市面上雖有針對腦電圖端的處理工具，但在心衝擊波偽跡校正上，除了參數設定較為複雜，且採半自動式抓取心跳複合波，以達到去除心跳干擾的目的，對於掃描時間較長的資料，此步驟的分析將費心耗時。因此，本研究欲改進心衝擊波偽跡校正方式，發展自動檢測心跳複合波演算法，同時也整合多項去除梯度偽跡之方法，加入fMRI頭動參數與主成分分析 (Principal Components Analysis, PCA) 提升偽跡去除效果，並且簡化校正所需參數設定。在MATLAB (MathWorks, Inc., MA, USA) 系統環境下，開發簡易操作之使用者介面，以此建立一個高效的全自動偽跡校正系統 (Automatic Artifact Correction System, AACS)。

與過去研究常使用的商用分析軟體Brain Vision Analyzer 2.0 (Brainproducts Gilching, Germany)、EEGlab的FMRIB工具箱 (Delorme and Makeig, 2004) 相比，本系統成功使MR (Magnetic Resonance) 梯度偽跡的基頻功率多衰減了4.269 %，並且全自動心跳偵測率提升至95 %以上。而心衝擊波偽跡的部分，偽跡殘留量多減少了1.442%，與心跳的相關性也由0.092降至0.073。同時，為了確保偽跡去除乾淨，神經電訊號也有成功被保留下來。我們利用Stroop任務誘發腦電活動，計算事件相關電位 (Event-Related Potential, ERP) 的訊號雜訊比 (Signal-to-Noise Ratio, SNR)，作為評估神經功能的替代指標。以掃描室外所收集到，不受環境干擾之腦電圖作為標準，與分別以Analyzer、FMRIB、AACS三種校正系統進行校正後的EEG做比較。結果發現本研究開發之AACS系統使校正後ERP的SNR，從原先的6.639、10.344提升至11.722，與在掃描室外收集到數據所得的12.378相比，差距大幅減少，成功提升了SNR。

最後，我們也嘗試以三個系統分別校正後的ERP成分，製作空間拓樸圖，觀察健康老化是否導致任務執行差異。結果顯示透過AACS系統去除偽跡的腦電圖，在空間拓樸圖的分布範圍更接近於掃描室外的結果。同時，對比年輕組於任務執行期間的枕葉活化表現，可發現老年組的激活腦區更多轉移至額、頂葉。此結果是以神經電訊號技術，更直接地再現過去研究中利用功能性磁振造影，所觀察到的大腦老化模型，衰老後前移 (Posterior-Anterior Shift in Aging, PASA)。

關鍵字

EEG-fMRI ； MR梯度偽跡 ； 心衝擊波偽跡 ； 腦波偽跡校正 ； Stroop任務 ； 事件誘發電位

並列摘要

In recent years, multimodal imaging technology is often used in the research of brain function, especially in the research field related to neurovascular coupling. It often integrates the results of different images and discusses its mechanism. The simultaneous acquisition of neural and vascular signals is very important for the study of this coupling phenomenon. Although the non-invasive synchronous scanning technology of functional Magnetic Resonance Imaging (fMRI) and electroencephalography (EEG) can simultaneously record neuroelectric signals and vascular signals induced by neural activities, it will seriously interfere with EEG data : (1) MR gradient artifacts (GA) (2) ballistocardiogram (BCG). Therefore, how to remove the interference signal is one of the core problems of the synchronization technology. The technical threshold for its signal processing has not been widely used in EEG-fMRI synchronization technology.

So far, although many correction methods have been proposed for EEG artifacts, there is no consensus on which method can obtain the best data quality after correction. Although there are processing tools for EEG on the market, in addition to the complex parameter setting, semi-automatic capture of QRS complex is adopted in order to remove heartbeat interference. For data with long scanning time, the analysis of this step will be time-consuming. Therefore, this study aims to improve the artifact correction method of BCG artifacts and develop an algorithm for automatic detection of heartbeat. At the same time, a number of methods for removing GA artifacts are also integrated, fMRI head motion parameters and Principal Components Analysis (PCA) are added to improve the artifact removal effect, and the parameter setting required for correction is simplified. In the MATLAB (MathWorks, Inc., MA, USA) system environment, a simple user interface is developed to establish an efficient Automatic Artifact Correction System (AACS).

Compared with the commercial analysis software Brain Vision Analyzer 2.0 (brainproducts gilching, Germany) and the FMRIB toolbox of EEGlab (Delorme and makeig, 2004), which are often used in the past research. The system successfully attenuated the fundamental frequency power of the GA by 4.269% and increased the automatic heartbeat detection rate to more than 95%. In the part of BCG artifact, the residual artifact decreased by 1.442%, and the correlation with heartbeat also decreased from 0.092 to 0.073. At the same time, in order to ensure the removal of artifacts, the neuroelectric signal has also been successfully preserved. We used the Stroop task to induce EEG activity and calculated the Signal-to-Noise Ratio (SNR) of Event-Related Potential (ERP) as an alternative index to evaluate neural function. The EEG collected outside the scanning room without environmental interference was taken as the standard, and compared with the EEG corrected by Analyzer, FMRIB and AACS respectively. The results show that the AACS system developed in this study improves the SNR of the corrected ERP from 6.639, 10.344 to 11.722. Compared with the 12.378 obtained outside the scanning room, the gap is greatly reduced and the SNR is successfully improved.

Finally, we also try to use the ERP components corrected by the three systems to make a spatial topology to observe whether healthy aging leads to differences in task execution. The results showed that the activation distribution range of the EEG that was removed by the AACS system was closer to the results outside the scanning room. In addition, compared with the activation of occipital lobe in the young group during task execution, it can be found that the activated brain areas in the old group are more transferred to frontal and parietal lobes. The result is a more direct reproduction of the model of brain aging observed in previous studies using fMRI, Posterior-Anterior Shift in Aging (PASA).

並列關鍵字

EEG-fMRI ； MR gradient artifacts (GA) ； ballistocardiogram (BCG) artifacts ； EEG artifact correction ； Stroop task ； Event-Related Potential (ERP)

參考文獻

[1] M. Ullsperger and S. Debener, Simultaneous EEG and fMRI: recording, analysis, and application. Oxford University Press, 2010.

[2] M. Moosmann et al., "Correlates of alpha rhythm in functional magnetic resonance imaging and near infrared spectroscopy," Neuroimage, vol. 20, no. 1, pp. 145-158, 2003.

[3] C. Mulert et al., "Single-trial coupling of EEG and fMRI reveals the involvement of early anterior cingulate cortex activation in effortful decision making," Neuroimage, vol. 42, no. 1, pp. 158-168, 2008.

[4] K. Krakow et al., "Spatio-temporal imaging of focal interictal epileptiform activity using EEG-triggered functional MRI," Epileptic disorders, vol. 3, no. 2, pp. 67-74, 2001.

[5] P. J. Allen, G. Polizzi, K. Krakow, D. R. Fish, and L. Lemieux, "Identification of EEG events in the MR scanner: the problem of pulse artifact and a method for its subtraction," Neuroimage, vol. 8, no. 3, pp. 229-239, 1998.