1、A,1,DTI數(shù)據(jù)分析及應(yīng)用,舒 妮 博士,北京師范大學(xué)認(rèn)知神經(jīng)科學(xué)與學(xué)習(xí)研究所,A,2,Page 2,內(nèi)容提綱,DTI的研究?jī)?nèi)容,DTI數(shù)據(jù)處理流程,DTI Studio,FSL: FMRIBs Diffusion Toolbox,A,3,Page 3,擴(kuò)散張量成像的研究?jī)?nèi)容,纖維跟蹤算法,基于DTI的應(yīng)用研究,A,4,Dxyxy Dyy Dyz = (v1 v2 v3) 0 2 0,擴(kuò)散張量的數(shù)學(xué)描述,D=,特征分解 特征值: 123 0 特征向量: vivj , i j Page 4,Dxx Dxy Dxz 1 0 0 v1 v2 Dxz Dyz Dzz 0 0 3 v3,A,5,Pag

2、e 5,確定性跟蹤算法,跟蹤終止條件,Mori et al., Ann Neurol, 1999,A,6,Page 6,確 定 性 跟 蹤 結(jié) 果,Catani et al, Brain, 2005,粗大的白質(zhì)纖維束,A,7,Uncertainty,Page 7,纖維走向的不確定性,Jones, MRM, 2003,Linearity,Bootstrap 方法,A,8,Page 8,概率跟蹤算法,Direction Uncertainty,DTI Noise,Partial Volume Effects,Slide from Tri Ngo,A,9,Page 9,概率跟蹤的方法,Non-par

3、ametric (model free) approaches,Bootstrap method,HARDI: Q-ball, DSI,Parametric approaches,Prior knowledge and models: Bayesian framework Probability density function (PDF): local, global,How to estimate the distribution of fiber,orientations within a voxel?,A,10,Page 10,概率跟蹤的思想 Reference: Behrens, T

4、.E. et al. Characterization and propagation of uncertainty,in diffusion-weighted MR imaging. Magn Reson Med 50, 1077-88 (2003).,A,11,Page 11,概 率 跟 蹤 結(jié) 果,Friman et al, IEEE TMI,2006,A,12,Page 12,概率跟蹤的優(yōu)點(diǎn)：,估計(jì)纖維走向的不確定性,一定程度上解決纖 維交叉問(wèn)題,研究FA較低的灰質(zhì)腦區(qū)之間的解剖連接 跟蹤結(jié)果對(duì)噪聲更穩(wěn)定,定量描述空間任意兩個(gè)體素之間的連接概率,概率跟蹤的缺點(diǎn)：,需要采集較多梯度方向的

5、DTI圖像 計(jì)算量大，耗時(shí),A,13,Page 13,Connectivity-based classification of thalamic voxels produces clusters,Behrens et al, Nature Neuroscience, 2003,A,14,Page 14,Improvements on the diffusion tensor model,single fibre,multiple fibres,Slide from Saad Jbabdi,A,15,Page 15,確定性跟蹤常用軟件:,DTI Studio, MedINRIA, 3D Slic

6、er等,概率跟蹤常用軟件：,FSL,http:/www.fmrib.ox.ac.uk/fsl/fdt/,A,16,Page 16,擴(kuò)散張量成像的研究?jī)?nèi)容,纖維跟蹤算法,基于DTI的應(yīng)用研究,A,17,Page 17,擴(kuò)散屬性測(cè)度,以上三種情況的 ADC = 0.7 x 10-3 mm2/s,A,18,Page 18,A,19,Page 19,A,20,Page 20,A,21,Page 21,基于擴(kuò)散屬性測(cè)度的臨床研究,基于全腦配準(zhǔn)的分析方法,基于體素的統(tǒng)計(jì)分析（VBA）,基于白質(zhì)骨架的空間統(tǒng)計(jì)分析（TBSS）,基于感興趣區(qū)的分析方法,手工畫(huà)感興趣區(qū)的方法 基于纖維重建的定量分析,A,22,P

7、age 22,Voxel-Based Analysis (VBA),VBM on FA (Ashburner, 2000; Rugg-Gunn, 2001) Strengths,Fully automated smallest systematic shifts between groups can be incorrectly interpreted as FA change,No objective way to choose smoothing extent (6, 8 or 10 mm?),A,23,Page 23,A,24,Page 24,A,25,Page 25,A,26,Page

8、 26,A,27,Page 27,A,28,Page 28,A,29,Page 29,A,30,Page 30,A,31,精神分裂癥患者的VBA分析 FA降低的腦區(qū)：, ,cerebral peduncle; frontal regions; inferior temporal gyrus; medial parietal lobes; hippocampal gyrus; Insula; right anterior cingulum bundle; right corona radiata Page 31,Hao et al. Neuroreport 2006,A,32,Page 32,T

9、ract-Based Spatial Statistics,(TBSS),Part of FSL software,(http:/www.fmrib.ox.ac.uk/fsl/tbss/index.html),Overcome the drawbacks in VBA method, such as alignment issue and smoothing issue Flowchart,A,33,Page 33,TBSS steps in detail:,preprocessing - create FA images from your diffusion study data,tbss

10、_1_preproc - prepare your FA data in your TBSS working directory in the right format,tbss_2_reg - apply nonlinear registration of all FA images into standard space,tbss_3_postreg - create the mean FA image and skeletonise it,tbss_4_prestats - project all subjects FA data onto the mean FA skeleton,st

11、ats (e.g., randomise) - feed the 4D projected FA data into GLM modelling and thresholding in order to find voxels which correlate with your model.,A,34,Page 34,Do cross-subject voxelwise stats on,skeleton-projected FA,A,35,Page 35,Fig. TBSS results from 15 MS patients. A,B: 3D surface renderings of

12、the mean FA skeleton. C: Yellow shows the where FA correlates negatively with EDSS disability score. D: Red as above. In C and D, green shows the mean FA skeleton, blue shows the group mean lesion distribution, and the background image is the MNI152.,Smith et al., NeuroImage, 2006,A,36,Page 36,Schol

13、z et al., Nature Neuroscience 2009,A,37,Page 37,TBSS data acquisition requirement: Voxel size should be less than 3 3 3 mm3.,At least one b = 0 should be acquired; ideally one b = 0 image for every eight diffusion- weighted images.,b-value should be at least 800 s/mm2.,At least six-gradient directio

14、ns must be acquired. it is better to use more unique sampling directions (with isotropic angular density18) than to obtain repeat samples of the same set of directions.,SNR in the diffusion-weighted images should be maximized . An example protocol that should lead to,sufficiently high SNR is having

15、b =1,000 s mm2, 24,diffusion-weighted images and SNR greater than 15 in the b = 0 image.,A,38,Page 38,The data should not be upsampled (e.g., through unfiltered zero-padding during reconstruction) if this is done in such a way as to introduce ringing into the data.,If multiple repeats of b = 0 or di

16、ffusion-weighted images are to be acquired, they must not be averaged on the scanner (as theymust be coregistered before averaging, and any risk of averaging the complex data should be avoided).,Fat saturation should be used whenever possible to remove signal from the scalp, which can disrupt signal

17、 in the brain owing to chemical shift or ghosting artifacts.,A vitamin capsule leftright marker (oil, not water) should be attached to the right side of the head to avoid any leftright ambiguities during data conversion and analysis.,A,39,Page 39,Computing equipment:,Unix-based computers. AppleMac (

18、running Mac OS X version 10.4 or higher) and PCs (running Linux flavors RedHat 9, Enterprise, FC4, Suse 9.0-9.3 or Debian),High RAM requirements (particularly if tens of subjects are used in a study), it is likely that the,computer will need to be 64 bit. The computer should have at least a 1 GHz CP

19、U clock, 1 GB RAM, 5 GB swap and 20 GB free hard disk space.,If multiple networked computers (or a computer cluster) are available, the registration steps can be parallelized, greatly reducing the total computation time.,A,40,Page 40,白質(zhì)纖維束的定量分析,FA: Left Cingulum (Red) Right Cingulum (Blue),Parameter

20、ization process,Gong et al, Human Brain Mapping, 2005,A,41,Page 41,同正常人相比，早期盲人的視放射白質(zhì)擴(kuò)散異常，F(xiàn)A值顯著,降低，ADC和23顯著升高。,早期盲人大腦白質(zhì)擴(kuò)散異常研究,Shu et al, Human Brain Mapping, 2009,A,42,Page 42,單張量模型的假設(shè)無(wú)法解決纖維交叉問(wèn)題,纖維跟蹤技術(shù)的準(zhǔn)確性缺乏嚴(yán)格的評(píng)價(jià)體系,擴(kuò)散張量成像的局限性,A,43,Page 43,內(nèi)容提綱,DTI的研究?jī)?nèi)容,DTI數(shù)據(jù)處理流程,DTI Studio,FSL: FMRIBs Diffusion Toolb

21、ox,A,44,Page 44,數(shù)據(jù)處理的 基本流程,A,45,Page 45,DWI from Scanner S0 S1 S2 S3 S4 S5,S6,A,46,Page 46,Preprocessing,DICOM data conversion,Image quality check,Eddy current correction,A,47,Page 47,內(nèi)容提綱,DTI的研究?jī)?nèi)容,DTI數(shù)據(jù)處理流程,DTI Studio,FSL: FMRIBs Diffusion Toolbox,A,48,Page 48,DTI Studio,/ D

22、ownload & Install User Manual Mailing list,A,49,Page 49,Launching the Program and Hardware Requirement,DtiStudio-latest-x86.exe for Windows system,More than 1GB RAM is recommended,A,50,Page 50,Main Functions,Image Viewer,Diffusion Tensor Calculation Fiber Tracking and Editing 3D Visualization,Image

23、File Management,ROI Drawing and Statistics,A,51,Page 51,How to do tensor calculation,and fiber tracking?,A,52,Page 52,E:workTrainingExampleData,Raw data: MRIcroN dcm2nii.exe,(.img, .hdr, .bvec, .bval),Eddy current correction: AIR,Tensor, FA, MD calculation: DTIstudio Fiber tractography: DTIstudio RO

24、I selection,A,53,Page 53,第一步：對(duì)原始DICOM數(shù)據(jù)進(jìn)行格式轉(zhuǎn)換。利用MRIcroN軟件 中的dcm2nii.exe工具，將DTI原始數(shù)據(jù)文件夾拖入，即可得到 DTI掃描的梯度編碼文件.bval和.bvec，以及轉(zhuǎn)換后的NIFTI格式 的圖像文件（Output Format選擇4D NIfTI hdr/img）。,A,54,Page 54,第二步：對(duì)DTI圖像進(jìn)行頭動(dòng)和渦流校正。打開(kāi)DTI studio, File - MRI View3D, 選中上一步得到的4D .img文件，Image Parameters中選擇Image File Format為 Analyze

25、，點(diǎn)擊OK，然后在Image面板Image Processing區(qū)域選擇Automatic Image Registration (AIR)，按圖3進(jìn)行設(shè)置，然后點(diǎn)擊OK，等圖像配準(zhǔn)完成后， 在Image面板的Orthogonal Views區(qū)域的文件下拉框中看到Air_開(kāi)頭的一系列文 件，為校正后的DTI圖像文件，點(diǎn)擊Save，將Air_開(kāi)頭的所有文件選中，選擇 Raw Data，保存為一個(gè)4D的.dat文件。,MRI View,3D參數(shù),A,55,Page 55,AIR的參數(shù) 設(shè)置,A,56,Page 56,頭動(dòng)和渦流校正后的DTI圖像保存,A,57,Page 57,第三步：張量解算以及F

26、A, ADC等擴(kuò)散指標(biāo)的計(jì)算。打開(kāi)DTI studio, File - DTI Mapping, 選擇Philips REC格式，Continue，按圖5進(jìn)行參數(shù) 設(shè)置， Add a file中選中上一步保存的4D .dat文件，點(diǎn)擊OK，在,DtiMap面板的Calculation區(qū)域選中Tensor, Color Map etc.（計(jì)算ADC值 選擇ADC-Map），根據(jù)圖像選擇噪聲水平，點(diǎn)擊OK，然后等DTI Studio算完后在Image面板的Orthogonal View區(qū)域可看到計(jì)算出來(lái)的各 種擴(kuò)散屬性文件。對(duì)于想要保存的文件，如FA, EigenVector-0，Color Ma

27、p-0等可以分別進(jìn)行Save（.dat格式），便于下一次查看和使用。,A,58,Page 58,DtiMap面板進(jìn)行張量解算,A,59,Page 59,各種擴(kuò)散屬性的顯示,A,60,Page 60,第四步：纖維跟蹤及可視化。第一種方法：基于前面步驟，在DtiMap面板的 Fiber Tracking區(qū)域點(diǎn)擊Fiber Tracking，然后進(jìn)行參數(shù)設(shè)置，點(diǎn)擊OK，就會(huì)進(jìn) 行基于全腦體素（FA0.2）的纖維重建；第二種方法：如果上一步已保存FA和 Eigen Vector-0文件，可重新打開(kāi)DTI Studio, File-Fiber-Tracking，選上FA- Map文件和Principle

28、 Vector文件，并進(jìn)行參數(shù)設(shè)置，點(diǎn)擊OK，就會(huì)進(jìn)行基于全 腦體素（FA0.2）的纖維重建。通過(guò)任何一種方法，算完后右下角會(huì)出現(xiàn)Fiber 面板，再此面板中可以對(duì)特定纖維束進(jìn)行顯示和編輯，并可以對(duì)纖維屬性進(jìn)行統(tǒng) 計(jì)分析。,A,61,Page 61,纖維跟蹤方法2,A,62,Page 62,纖維跟蹤的參數(shù)設(shè)置,A,63,Page 63,重建纖維束的可視化,A,64,Page 64,A,65,Page 65,Major white matter tracts,Reference: Wakana S, Caprihan A, Panzenboeck MM, et al., Reproducibility of quantitative tractography methods applied to cerebral white matter. Neuroimage, 2007, 36(3): 630-644.,A,66,Page 66,纖維屬性 的統(tǒng)計(jì)分 析,A,67,Page 67,New Modules,ROIEditor,ROI drawing, s