| 摘要: | 近年來，靜息狀態功能性磁振造影(Resting-state Functional Magnetic Resonance Imaging, RS-fMRI)的研究被廣泛的運用在心理及臨床領域，因此受到越來越多的關注，也成為當今神經科學領域研究的熱門議題之一。即使相關研究逐年增加，靜息狀態功能性磁振造影之機制尚未明朗且仍需投入大量研究。雖然科學家並沒有完全了解靜息態功能性磁振造影的訊號源，但是基於血氧濃度依賴(Blood Oxygen Level-Dependent, BOLD)訊號的低頻波動可反映神經自發性的同步活化現象，因此提供我們探索大腦運作謎團的契機。然而自發性的波動非常微小，極易被系統雜訊及生理雜訊所影響。因此若能降低來自內在及外在因素所造成的雜訊，其大腦連結的特性將會更為顯著。所以，本文嘗試使用三種不同的雜訊移除方法並探討其用於靜息狀態功能性磁振造影分析之效果。
本研究於兩台3 Tesla (T)磁振造影儀(型號分別為Siemens Trio及Skyra)收集靜息態的資料，並在收集的同時量測呼吸及心跳的訊號。利用(1) 脈衝響應分析 (IRF) 、(2) RETROICOR及(3) 脈衝響應結合RETROICOR (IRFRETRO)來去除生理雜訊。目前靜息狀態功能性磁振造影信號並沒有特定的指標可衡量資料的穩定性，因此本論文提出以像素為基礎之五種指標：低頻振幅(ALFF)、低頻振幅比例(fALFF)、 區域同質性(ReHo)、空間及時間訊雜比(spatial and temporal SNR)，來估測資料的品質，並針對不同腦區：全腦(Whole Brain,WB)、前額皮層(Orbitofrontal Cortex,OFC)、視覺皮質區(Visual Cortex,VC)、杏仁核(Amygdala,AMY)來探討腦區間的特性。
結果顯示生理雜訊對於五種指標之影響甚小，無論使用何種去除生理雜訊之方法皆無顯著差異；而在腦區間的比較，視覺皮質區相較於其他區域有著較強烈的區域同質性；在磁振造影儀器之間的比較，Skyra比起Trio有較高的空間與時間訊雜比，意即全腦及視覺皮質區域的靜息狀態功能性磁振造影訊號在Trio上較易受到系統雜訊之影響；而Skyra則是在空間解析度方面有較大的受試者間差異。結果表示，從靜息狀態功能性磁振造影訊號移除生理雜訊於觀察上述五種指標分析時並非必須執行之程序。這項觀察結果可能意味著生理訊號（呼吸、心跳）也包含在靜息狀態功能性磁振造影訊號之一部份，單純視其為雜訊可能並不適當，這方面還需要更進一步的研究探討。
Recently, the investigations of resting-state functional magnetic imaging (RS-fMRI) have attracted attention to the neuroscience field for its wide application on psychological exploration and the clinical pathologies. Beyond its proliferation, the mechanism of RS-fMRI remains unknown and under intensive investigations. Although scientists do not clarify the RS-fMRI signal sources, the blood oxygen level-dependent (BOLD)-based RS-fMRI datasets could be influenced by pre-defined noise types, such as instrumental and physiological noise. Therefore, it is claimed that if the noise of RS-fMRI can be reduced, the significance of the functional connectivity is supposed to increase. To this point, we tested three different noise removal methods to investigate their effectiveness on RS-fMRI analyses.
We collected RS-fMRI data from two 3 Tesla (T) scanners, and measured the physiological monitoring unit, including respiration and cardiac pulsation signals, simultaneously. For evaluating the effect of physiological noise, we used three ways to estimate the physiological noise: (1) impulse response function (IRF), (2) RETROICOR and (3) the combination of IRF and RETROICOR (IRFRETRO). Then we compared the effect of these methods on five voxel-based RS-fMRI indices (ALFF, fALFF, ReHo, spatial and temporal SNR), including different brain regions: whole brain (WB), orbitofrontal cortex (OFC), visual cortex (VC) and amygdala (AMY) to further evaluate the regional characteristics.
The result showed that the physiological noise has minimal effect on the five RS-fMRI indices, no matter which type of noise removal methods (IRF, RETROICOR or IRFRETRO) was applied. In the regional comparison,, only VC had unique characteristics compared to other regions, especially for ReHo. In the scanner comparison, Skyra possessed higher spatial and temporal SNR than Trio. More specifically, the RS-fMRI signals of WB and VC were dominated by instrumental noise in Trio scanner, whereas the Skyra scanner had higher between-subject variability on spatial SNR. In conclusion, inclusion of the noise removal on RS-fMRI signal may be unnecessary for the five indices. The observations implied that whether we should consider the physiological noise (respiration and cardiac pulsation) as a ‘pure noise’ in RS-fMRI signal remains to be discussed. |
