以功能性定量磁化率影像與腦磁圖探討負活化血氧相依機制

Abstract

功能性磁振造影(functional MRI, fMRI)為一項非侵入性的神經影像技術，以血氧濃度相依對比機制(Blood Oxygenation Level Dependent, BOLD)為基礎，量測神經活動所引發的血氧濃度變化，能夠定位大腦功能區域，於神經科學領域中有著相當廣泛的應用。然而BOLD機制並不完整，目前科學家們使用神經血管耦合來解釋正活化血氧相依反應，但負活化血氧相依反應(Negative BOLD Response, NBR)的成因卻不清楚，主要原因在過去NBR假說無法一統多年來研究者觀測到的fMRI的現象，目前為止NBR有三種假說 : (1)血管性成因的血液竊取假說(2)神經血管混和性的神經血氧不同步假說(3)神經性成因的神經抑制假說。每個假說有各自的生理機制與證據支持，導致負活化血氧相依機制尚未有定論。本篇論文探討了複雜認知任務在預設模式網絡(Default mode network, DMN) 誘發之NBR，使用功能性定量磁化率影像(functional QSM, fQSM)技術定量血氧濃度變化之訊息，並使用腦磁圖(magnetoencephalography, MEG) 觀察負活化反應區域的神經活動現象，藉此來探討NBR背後的機制，尤其利用神經資訊來討論神經抑制假說。此外，fQSM靈敏度會受到磁化率影像計算過程中產生的假影及雜訊影響，本研究的另一目的在於提升fQSM偵測率，利用以腦脊隨液為零基準之正則化方法進行優化。 本研究收取18位受試者之功能性磁振影像，實驗採用2-back工作記憶任務誘發負活化血氧相依反應。時間序列相位影像進行相位展開、去除背景磁場以及腦脊隨液為零基準之正則優化磁化率影像分析流程，隨後進行功能性影像前處理及統計分析計算大腦活化區域。腦磁圖實驗採用相同任務，收集之腦磁訊號使用時頻分析及溯源分析觀察頻帶功率變化及活動源，利用頻帶功率變化具有神經興奮或抑制的特性來探討負血氧機制。為了連結fQSM與腦磁圖資料，以fQSM活化區域的峰值建立ROI，圈選18位受試者fQSM磁化率變化與MEG溯源分析之功率，進行皮爾森相關性分析，以及使用餘弦相似性計算fQSM活化圖譜與MEG功率圖譜的相似性。 優化後之fQSM於腦室及大腦邊界處減少陰影假影，降低時間序列QSM影像的變異性，使fQSM能夠偵測多出1.54倍的共同體素比例。fQSM與腦磁圖的相關性分析中，於DMN的楔前葉觀察到NBR 與Alpha功率成正相關(r=0.51,p=0.04)，且與其他功率相比有最高的餘弦相似性，表明了此區域的NBR可能與神經抑制有關，並推測執行任務時DMN的負活化可能參與抑制非任務相關資訊，使大腦可以專注執行記憶任務。 綜上所述，藉由fQSM及腦磁圖技術，本研究在負活化血氧相依領域中，為神經抑制假說提出了初步的論證，此應用顯示優化後的fQSM將能發揮其定量及去除非局部效應之潛力，配合腦磁圖觀察神經活動，彰顯了神經與血氧技術應用於長期、介入性臨床研究的可能性，相信對神經科學領域之精準醫療有極大助益。

Functional magnetic resonance imaging (fMRI) is a non-invasive MRI technique that utilizes the blood oxygenation-level dependent (BOLD) mechanism to measure the changes in blood oxygen concentration. It discovers the function of the brain and widely applies in neuroscience. Current researchers can explain the positive BOLD response (PBR) mechanism with the neurovascular coupling principle. However, the cause of the negative BOLD response (NBR) is still unclear because the fMRI phenomena regarding NBR observed in the past cannot be fully explained by the three hypotheses, which are (1) the vascular hypotheses “the blood stealing” (2) the mix hypotheses “mismatch of hemodynamic response function and neural activity” (3) the neural hypothesis “the neural inhibition.” All three hypotheses have their physiological basis and evidence, so the mechanism of NBR remains a mystery. This thesis explored the NBR by applying two quantitative techniques, functional quantitative susceptibility mapping (fQSM) and magnetoencephalography (MEG), to quantify the changes in blood oxygen concentration and nerves in the default mode network (DMN) region which is usually observed the negative BOLD response during complex cognitive tasks. Additionally, the sensitivity of fQSM suffers from QSM processing especially solving an ill-posed deconvolution. Another purpose of this study is to improve the detection rate of fQSM by using a regularization method with cerebrospinal fluid (CSF) as the zero reference. 18 healthy participants were recruited and they completed a 2-back working memory task during fMRI. Time-series fMRI phase images were calculated by QSM processing including phase unwrapping, background field removal and dipole inversion using cerebrospinal fluid regularization as a zero reference to generate susceptibility images. Then, images were computed by functional and statistical analysis to produce the brain activation map. The participants also performed the MEG experiment adapting the same 2 back task to induce brain neural activation. Brain signal were recorded and analyzed using time frequency analysis and source analysis to generate time-frequency response and band power map. The NBR hypotheses were explored by using the frequency band power change with the characteristics of neural excitation or inhibition. To fuse the fQSM and MEG data, we established the ROI in the peak value of fQSM activation area. The susceptibility change and band power change in the ROI were selected to calculate Pearson correlation in 18 subjects. And we also evaluated the similarity between fQSM activation map and MEG band power map using the cosine similarity evaluation. The optimized fQSM suppressed shadowing artifacts at the ventricles and brain boundaries and reduced the variability by enhancing 1.54 times common voxel ratios compared with original processing. In the correlation analysis, the susceptibility change in NBR observed in the precuneus of DMN is positively correlated with alpha power (r=0.51, p=0.04), and the highest cosine similarity compared with other band power also observed in the region. The results indicated that the NBR in precuneus may be related to the neural inhibition hypothesis. It suppressed the irrelevant information, so that the brain can focus on memory tasks. In summary, this study applied fQSM and MEG to explore the NBR mechanism and provided preliminary evidence for the neural inhibition hypothesis. This application shows that the optimized fQSM combined with MEG has potential for application to long-term and interventional clinical research to observe the blood oxygen and neural activity, which is believed to be of great benefit to precision medicine in neuroscience.