評估第五腦室及中顱窩蜘蛛膜囊腫造成特定神經纖維束擴散係數之取樣誤差

Evaluation of errors in tract-specific sampling of diffusion index due to the cavum septum pellucidum and middle cranial fossa arachnoid cyst

翁振翔(Chen-Hsiang Weng)

指導教授 ： 曾文毅

國立臺灣大學/醫學院/醫療器材與醫學影像研究所/碩士(2018年)

https://doi.org/10.6342/NTU201803924

摘要

研究目的
大腦中的異常結構，如第五腦室(cavum septum pellucidum)及顳窩蜘蛛膜囊腫(middle cranial fossa arachnoid cyst)，因其好發率低(第五腦室約15-20%;顳窩蜘蛛膜囊腫約1%)，導致包含此異常結構的磁振造影影像在分析過程中易被忽略。近年來，對於腦部異常結構的存在是否影響影像分析的結果逐漸被討論，但目前仍沒有文獻探討相關的議題。因此，本研究使用擴散頻譜造影影像(diffusion spectrum imaging)評估第五腦室及顳窩蜘蛛膜囊腫造成影像對位或擴散係數計算的影響，並提出相對應的檢核標準。

研究方法
實驗一:此研究共使用45位有第五腦室的患者及45位腦部結構正常的控制組，並分析每一個人的T1權重影像及擴散頻譜造影影像。首先，利用T1影像測量第五腦室的大小，將其分為三組(大於10 毫米為大尺寸組，介於5-10毫米為中尺寸組，小於5毫米則為小尺寸組)，並且使用Dice係數(Dice coefficient)探討不同大小的第五腦室對影像組織分割(segmentation)的影響，再透過全腦神經束之自動化分析(Tract-based automatic analysis, TBAA)擷取全腦主要76條神經纖維束之非等向性指標(generalized fractional anisotropy, GFA)，藉此探討特定白質神經束結構與第五腦室大小之相關性。
實驗二:此研究共使用10位有顳窩蜘蛛膜囊腫的患者及30位腦部結構正常的控制組，分析每一個人的T1權重影像及擴散頻譜造影。在T1影像中，執行區域成長(region growing)演算法，計算出顳窩蜘蛛膜囊腫的體積，再運用全腦白質神經纖維束分析方式及功能性差異(functional difference)探討左右兩側的白質神經束結構受顳窩蜘蛛膜囊腫體積之影響。

研究結果與討論
實驗一:影像組織分割的結果顯示大尺寸組第五腦室與控制組呈現顯著的差異，將造成影像對位的錯誤，而中、小尺寸組則沒有顯著的差異。多元回歸分析(multiple regression analysis)的結果顯示有七條白質神經纖維束與第五腦室的大小呈現顯著相關。兩條連絡纖維(兩側的終紋(stria terminalis))跟第五腦室呈現顯著負相關，且第五腦室影響終紋的路徑，造成擴散係數的取樣誤差。五條投射纖維(右側連結手部的皮質脊髓束(corticospinal tracts (CST) of the hand)，右側連結軀幹的皮質脊髓束(CST of the trunk)，兩側連結嘴部的皮質脊髓束(CST of the mouth)及右側連接中央前迴的額葉紋狀體神經路徑(frontostriatal tract (FS) of precentral gyrus))與第五腦室呈現顯著正相關，此結果可能為第五腦室壓迫週遭的大腦白質使得GFA上升。
實驗二:以控制組每一條神經束之功能性差異的平均值加兩倍標準差作為閾值，若顳窩蜘蛛膜囊腫患者的特定白質神經束大於閾值視為有顯著差異。結果顯示若顳窩蜘蛛膜囊腫大於21890立方毫米，兩側的下縱束(inferior longitudinal fasciculus)及兩側的顳葉端聯合神經束(callosal fibers connecting the temporal poles)有顯著性差異; 若顳窩蜘蛛膜囊腫大於39940立方毫米，兩側的鉤狀束(uncinate fasciculus) 有顯著性差異。此結果表示患側的白質神經束受顳窩蜘蛛膜囊腫影響使得GFA下降，導致兩側神經束的功能性差異顯著大於閾值。

總結
研究結果顯示第五腦室及顳窩蜘蛛膜囊腫的存在確實會導致擴散係數的計算誤差，且與其大小及體積有關。因此本研究針對這兩種異常結構各提出相對應的檢核標準。實驗一:於統計分析中，須將第五腦室的大小作為共變量。若第五腦室大於10mm，影像分析的結果須排除兩側的終紋。實驗二:若顳窩蜘蛛膜囊腫的體積大於21890 mm³ (以球體體積來計算，半徑約為17.3mm)，兩側的下縱束及兩側的顳葉端聯合神經束須被排除；若顳窩蜘蛛膜囊腫的體積大於39940 mm³ (半徑約為21.2mm)，兩側的鉤狀束也須被排除。

關鍵字

第五腦室 ； 顳窩蜘蛛膜囊腫 ； 擴散頻譜造影 ； 大腦白質神經纖維束 ； 全腦白質神經纖維自動化分析

並列摘要

Introduction
The abnormal brain structure, including cavum septum pellucidum (CSP) and middle cranial fossa (MCF) arachnoid cyst, were seen in 15-20% and up to 1.1% in the CSP and MCF arachnoid cyst, respectively. The prevalence is low so that we are easy to include the MRI data without scrutiny and used for analysis. In recent years, it becomes more and more attention on this issue, which affects the errors of the image registration or diffusion measurement in the neuroimaging analysis. However, there is no study or researches to explore the issues. Therefore, this study aimed to identify the errors of the tissue segmentation in CSP and white matter reconstruction caused by different sizes of the CSP and the MCF arachnoid cyst, and propose criteria for using this data in neuroimaging analysis.

Materials and Methods
Experience 1: Forty-five subjects with the CSP and forty-five age- and sex-matched controls without the CSP were recruited in the study. DSI data and T1-weighted imaging were used for imaging analysis. First, we measured the size of the CSP on each T1-weighted image, and divided these subjects into three groups according to the size of CSP (Enlarged-size: above 10 mm, n = 14; Medium-size: 5-10 mm, n = 16; Small-size: below 5 mm, n = 15). Then, we used Dice coefficient to quantify the errors of the image registration caused by the CSP in T1W segmentation. Second, we used whole brain tract-based automatic analysis to obtain a 2D connectogram for each DSI dataset. Generalized fractional anisotropy (GFA) profiles of 76 white matter tract bundles provided by connectogram used to explore the correlation with the size of the CSP
Experience 2: Ten subjects with the middle cranial fossa arachnoid cyst and thirty age- and sex-matched controls without the MCF arachnoid cyst were recruited in the study. DSI data and T1-weighted imaging were used for imaging analysis. The subjects with MCF arachnoid cyst was segmented clearly by region growing algorithm in each T1-weighted images. Then, we performed diffusion spectrum imaging and the functional difference to investigate the relations between white matter microstructures and the volume of the MCF arachnoid cyst.
Results and Discussions
Experience 1: Dice coefficient evaluated in 3 groups showed significantly lower Dice coefficients (p<.001) in the enlarged-size of the CSP, but there was no significant difference (p=.631 and p=.317) in medium and small CSP subjects. Multiple regression analysis showed that the seven tracts were significantly associated with the size variable. The mean GFA values of the bilateral stria terminalis were negatively correlated with the size of the CSP, and the paths of the stria terminalis were deviated by the CSP, leading to the errors of diffusion measurement. The other five tracts, right CST of the hand, right CST of the trunk, bilateral CST of the mouth and left FS of the precentral gyrus, showed significant differences with control groups, and the GFA values of the CSP groups showed slightly higher in all steps than control groups. We conjectured that these white matter tracts are projection fibers in close proximity to the lateral ventricles, and so they might be compressed by the CSP, leading to the increased GFA.
Experience 2: Apply two standard deviations as threshold, we found that the FD values of the bilateral inferior longitudinal fasciculus (ILF) and callosal fibers (CF) of temporal pole were significantly higher than the control group in the largest and second largest volume of the MCF arachnoid cyst, which was 39940 and 21890 mm3; the FD values of the bilateral uncinate fasciculus (UF) were significantly higher than the control group in the largest volume of the MCF arachnoid cyst, which was 39940 mm3. The higher FD values in ILF, UF, and CF of temporal pole meant that the difference between left and right tracts was greater, which affected by the MCF arachnoid cyst.

Conclusion
Our findings imply that the existence of the CSP and MCF arachnoid cyst will lead to errors in image registration and diffusion measurement, and the size of CSP and the MCF arachnoid cyst affects the GFA values in the adjacent tracts. Thus, we propose criteria as follows:
Experience 1: we suggested that the size of the CSP should be taken into consideration as a covariate in statistical analysis, and bilateral stria terminalis should be excluded from the analysis if the CSP is larger than 10 mm.
Experience 2: we suggested that the bilateral ILF and CF of temporal pole should be excluded from the analysis if the cyst is larger than 21890 mm³, roughly R=17.3mm, and the bilateral UF should be excluded from the analysis if the cyst is larger than 39940 mm³, roughly R=21.2mm.

並列關鍵字

Cavum septum pellucidum ； Middle cranial fossa arachnoid cyst ； Diffusion spectrum image ； Tract-based automatic analysis ； White matter tract

參考文獻

Wedeen, V. J., Hagmann, P., Tseng, W. Y., Reese, T. G., & Weisskoff, R. M. (2005). Mapping complex tissue architecture with diffusion spectrum magnetic resonance imaging. Magn Reson Med, 54(6), 1377-1386. doi:10.1002/mrm.20642

Wedeen, V. J., Wang, R. P., Schmahmann, J. D., Benner, T., Tseng, W. Y., Dai, G., . . . de Crespigny, A. J. (2008). Diffusion spectrum magnetic resonance imaging (DSI) tractography of crossing fibers. Neuroimage, 41(4), 1267-1277. doi:10.1016/j.neuroimage.2008.03.036

Assaf, Y., Ben-Sira, L., Constantini, S., Chang, L. C., & Beni-Adani, L. (2006). Diffusion tensor imaging in hydrocephalus: initial experience. AJNR Am J Neuroradiol, 27(8), 1717-1724.

Born, C. M., Meisenzahl, E. M., Frodl, T., Pfluger, T., Reiser, M., Moller, H. J., & Leinsinger, G. L. (2004). The septum pellucidum and its variants. An MRI study. Eur Arch Psychiatry Clin Neurosci, 254(5), 295-302. doi:10.1007/s00406-004-0496-z

Callaghan, P. T., Coy, A., Macgowan, D., Packer, K. J., & Zelaya, F. O. (1991). Diffraction-Like Effects in Nmr Diffusion Studies of Fluids in Porous Solids. Nature, 351(6326), 467-469. doi:DOI 10.1038/351467a0