探討角度閾值對以神經束為基礎之擴散張量影像分析的影響

Abstract

研究目的： 磁振神經束造影(Magnetic Resonance Tractography)是透過擴散影像以觀察水分子在大腦中的擴散特性，進而以非侵入性的方式建立大腦白質微結構。透過三維神經纖維束的視覺化對於神經科學研究以及臨床應用上極有幫助。此神經束追蹤方法中包含演算方式、閾值設定等多項參數可以調整，本研究以定量分析探討其一重要參數：角度閾值(Angular threshold, α)，如何影響神經束之擴散性指標與其神經纖維束結構，並加以了解基於模板的神經束分析可能因存在病變或神經束重組時而忽略了結構變異的可能性。 此研究中，透過擴散性指標與白質纖維結構來觀察改變「角度閾值」後的變化與差異。然而目前對這一技術問題的系統性評估仍然很少。本研究旨在通過量化角度閾值的影響作為闡述主題。同時，過去的文獻中經常使用正常受試者之模板方式進行研究分析，此研究另一目的為探討白質纖維可能在患者上的結構變異而具有更多可變走向，因此在使用白質模板分析數據時應更加謹慎。 研究方法： 我們招收20名健康老人與16名阿茲海默症患者。此研究於3T 磁振造影儀(Tim Trio，Siemens)上進行36位受試者之擴散張量影像掃描，觀察大腦白質神經纖維束之水分子擴散特性，計算水分子在白質神經纖維中之擴散性指標。腦部的神經纖維可按照其聯絡的系統分為投射纖維(Projection fibers)、聯合纖維(Association fibers)以及連合纖維(Commissural fibers)三類，此研究使用DSI Studio軟體對於其六種神經纖維束進行分析，平均每條神經纖維束之路徑上的擴散性指標，並且使用MATLAB重新描繪不同神經纖維束之結構變化。同時，透過標準化(Normalization)擴散性指標，描繪出控制組之群體中的平均數值變化趨勢。接著使用T檢定檢驗調整「角度閾值」後的數值變化與差異。更進一步地使用共變異數分析(Analysis of Covariance, ANCOVA)探討角度閾值對於控制組與病人組之間臨床差異的影響。 研究結果與討論： 研究結果顯示角度閾值確實造成影響。首先結構上，可以在皮質脊髓束(Corticospinal tract, CST)中發現，當訊號雜訊比(Signal-to-Noise Ratio, SNR)到達一定水準時，調高角度閾值隨之能夠追蹤出錐體交叉的發生。而在擴散性指標上，提高角度閾值會造成非等向性擴散指標(Fractional Anisotropy, FA)的下降和平均擴散性指標(Mean Diffusivity, MD)的上升。此外，經由共變異數分析，控制組和病人組在丘腦前輻射(Anterior Thalamic Radiation, ATR)、鉤束(Uncinate Fasciculus, UF)及小鉗纖維束(Forceps minor, F-minor)的擴散性指標之顯著差異會因著調整閾值設定而有所變化。且當圈選相同區域進行神經纖維追蹤時，小鉗纖維束在控制組與病人組之間具有明顯不同的結構走向變異。 結論： 本研究結果顯示角度閾值的選擇對於神經纖維追蹤的擴散性指標和白質結構造成變化。而這些影響可能混淆臨床解釋與解剖意義，同時也需要將訊雜比和感興趣區域圈選(Regions of interest, ROIs)一同考量進去。角度閾值會影響包含的纖維組成和神經束之擴散性指標，因此在進行交叉引用和臨床解釋時需更為謹慎。

Introduction: Tractography is a non-invasive technique used to establish the microstructure of white matter in the brain through MR diffusion imaging. The visualization of nerve fiber bundles in three dimensions is extremely helpful for neuroscience research and clinical applications. This tracking method has many parameters that can be adjusted including calculation method and threshold setting. This study aims to quantitatively investigate how angular threshold (α) may affect tract-based measurement of diffusion metrics and observe the microstructure of white matter tracts. Systematic assessment of this technical issue remains scarce. The purpose of this study is to elucidate the issue by quantifying the effect of angular threshold (α). Furthermore, template-based analysis has been popular in neuroimaging studies, assuming that structural variabilities in the study population are largely removable by normalization, which can be violated when there is lesion and/or reorganization. This study also wants to demonstrate that white matter fibers may be rerouted in dementia patients. Materials and methods: Twenty healthy subjects and sixteen dementia patients were included and all gave written informed consent before participation. MR exams were conducted on a 3-Tesla whole-body system. The nerve fibers in the brain can be divided into three types: projection fibers, association fibers and commissural fibers according to their communication systems. This study analyzed six fiber tracts by using DSI Studio software for analysis, and then using MATLAB to re-delineate the structural changes of different fiber tracts. By normalizing the diffusion metrics, the trend of mean FA and MD value of obtained tracts was plotted. Paired t test was then used to examine differences in FA and MD values after adjusting the angular threshold (α) for the control group. Moreover, analysis of covariance (ANCOVA) was used to explore whether the clinical differences between control subjects and dementia patients were varied due to the adjustment of the angular threshold, which led to the significant difference between two groups. Results and Discussions: The results of this research showed that the angular threshold (α) make an impact. First, structurally, it can be found that in the corticospinal tract, the occurrence of pyramidal decussation can be tracked as the threshold is increased. In addition, our data demonstrated that the forceps minor had more variable pathways and the characteristic U-shape became less recognizable in dementia patients, which would be missed with template-based analysis. On the diffusion metrics, increasing the angular threshold would cause a decrease in FA value and an increase in the MD value. Besides, significant differences of the diffusion metrics on anterior thalamic radiation, uncinate fasciculus, and forceps minor between control subjects and dementia patients varied by adjusting the threshold settings. And when the same ROIs were picked for fiber tracking, the fiber tract structure of F-minor had obviously different pathways between two groups. Conclusion: The present study results supported that α has a non-trivial effect on the results of tract-based analysis. The effect can be anatomically meaningful and should be considered along with SNR and ROIs. Angular threshold can affect the included fiber composition and the values of diffusion metrics obtained with tract-based analysis, and thus needs to be considered when making cross-reference and clinical interpretation.