DTI数据分析及应用.ppt

DTI数据分析及应用,舒妮博士,北京师范大学认知神经科学与学习研究所,Page2,内容提纲,DTI的研究内容,DTI数据处理流程,DTIStudio,FSL:FMRIBsDiffusionToolbox,Page3,扩散张量成像的研究内容,纤维跟踪算法,基于DTI的应用研究,DxyxyDyyDyz=(v1v2v3)020,扩散张量的数学描述,D=,特征分解特征值:1230特征向量:vivj,ijPage4,DxxDxyDxz100v1v2DxzDyzDzz003v3,Page5,确定性跟踪算法,跟踪终止条件,Morietal.,AnnNeurol,1999,Page6,确定性跟踪结果,Catanietal,Brain,2005,粗大的白质纤维束,Uncertainty,Page7,纤维走向的不确定性,Jones,MRM,2003,Linearity,Bootstrap方法,Page8,概率跟踪算法,DirectionUncertainty,DTINoise,PartialVolumeEffects,SlidefromTriNgo,Page9,概率跟踪的方法,Non-parametric(modelfree)approaches,Bootstrapmethod,HARDI:Q-ball,DSI,Parametricapproaches,Priorknowledgeandmodels:BayesianframeworkProbabilitydensityfunction(PDF):local,global,Howtoestimatethedistributionoffiber,orientationswithinavoxel?,Page10,概率跟踪的思想Reference:Behrens,T.E.etal.Characterizationandpropagationofuncertainty,indiffusion-weightedMRimaging.MagnResonMed50,1077-88(2003).,Page11,概率跟踪结果,Frimanetal,IEEETMI,2006,Page12,概率跟踪的优点：,估计纤维走向的不确定性,一定程度上解决纤维交叉问题,研究FA较低的灰质脑区之间的解剖连接跟踪结果对噪声更稳定,定量描述空间任意两个体素之间的连接概率,概率跟踪的缺点：,需要采集较多梯度方向的DTI图像计算量大，耗时,Page13,Connectivity-basedclassificationofthalamicvoxelsproducesclusters,Behrensetal,NatureNeuroscience,2003,Page14,Improvementsonthediffusiontensormodel,singlefibre,multiplefibres,SlidefromSaadJbabdi,Page15,确定性跟踪常用软件:,DTIStudio,MedINRIA,3DSlicer等,概率跟踪常用软件：,FSL,http:/www.fmrib.ox.ac.uk/fsl/fdt/,Page16,扩散张量成像的研究内容,纤维跟踪算法,基于DTI的应用研究,Page17,扩散属性测度,以上三种情况的ADC=0.7x10-3mm2/s,Page18,Page19,Page20,Page21,基于扩散属性测度的临床研究,基于全脑配准的分析方法,基于体素的统计分析（VBA）,基于白质骨架的空间统计分析（TBSS）,基于感兴趣区的分析方法,手工画感兴趣区的方法基于纤维重建的定量分析,Page22,Voxel-BasedAnalysis(VBA),VBMonFA(Ashburner,2000;Rugg-Gunn,2001)Strengths,FullyautomatedsmallestsystematicshiftsbetweengroupscanbeincorrectlyinterpretedasFAchange,Noobjectivewaytochoosesmoothingextent(6,8or10mm?),Page23,Page24,Page25,Page26,Page27,Page28,Page29,Page30,精神分裂症患者的VBA分析FA降低的脑区：,cerebralpeduncle;frontalregions;inferiortemporalgyrus;medialparietallobes;hippocampalgyrus;Insula;rightanteriorcingulumbundle;rightcoronaradiataPage31,Haoetal.Neuroreport2006,Page32,Tract-BasedSpatialStatistics,(TBSS),PartofFSLsoftware,(http:/www.fmrib.ox.ac.uk/fsl/tbss/index.html),OvercomethedrawbacksinVBAmethod,suchasalignmentissueandsmoothingissueFlowchart,Page33,TBSSstepsindetail:,preprocessing-createFAimagesfromyourdiffusionstudydata,tbss_1_preproc-prepareyourFAdatainyourTBSSworkingdirectoryintherightformat,tbss_2_reg-applynonlinearregistrationofallFAimagesintostandardspace,tbss_3_postreg-createthemeanFAimageandskeletoniseit,tbss_4_prestats-projectallsubjectsFAdataontothemeanFAskeleton,stats(e.g.,randomise)-feedthe4DprojectedFAdataintoGLMmodellingandthresholdinginordertofindvoxelswhichcorrelatewithyourmodel.,Page34,Docross-subjectvoxelwisestatson,skeleton-projectedFA,Page35,Fig.TBSSresultsfrom15MSpatients.A,B:3DsurfacerenderingsofthemeanFAskeleton.C:YellowshowsthewhereFAcorrelatesnegativelywithEDSSdisabilityscore.D:Redasabove.InCandD,greenshowsthemeanFAskeleton,blueshowsthegroupmeanlesiondistribution,andthebackgroundimageistheMNI152.,Smithetal.,NeuroImage,2006,Page36,Scholzetal.,NatureNeuroscience2009,Page37,TBSSdataacquisitionrequirement:Voxelsizeshouldbelessthan333mm3.,Atleastoneb=0shouldbeacquired;ideallyoneb=0imageforeveryeightdiffusion-weightedimages.,b-valueshouldbeatleast800s/mm2.,Atleastsix-gradientdirectionsmustbeacquired.itisbettertousemoreuniquesamplingdirections(withisotropicangulardensity18)thantoobtainrepeatsamplesofthesamesetofdirections.,SNRinthediffusion-weightedimagesshouldbemaximized.Anexampleprotocolthatshouldleadto,sufficientlyhighSNRishavingb=1,000smm2,24,diffusion-weightedimagesandSNRgreaterthan15intheb=0image.,Page38,Thedatashouldnotbeupsampled(e.g.,throughunfilteredzero-paddingduringreconstruction)ifthisisdoneinsuchawayastointroduceringingintothedata.,Ifmultiplerepeatsofb=0ordiffusion-weightedimagesaretobeacquired,theymustnotbeaveragedonthescanner(astheymustbecoregisteredbeforeaveraging,andanyriskofaveragingthecomplexdatashouldbeavoided).,Fatsaturationshouldbeusedwheneverpossibletoremovesignalfromthescalp,whichcandisruptsignalinthebrainowingtochemicalshiftorghostingartifacts.,Avitamincapsuleleftrightmarker(oil,notwater)shouldbeattachedtotherightsideoftheheadtoavoidanyleftrightambiguitiesduringdataconversionandanalysis.,Page39,Computingequipment:,Unix-basedcomputers.AppleMac(runningMacOSXversion10.4orhigher)andPCs(runningLinuxflavorsRedHat9,Enterprise,FC4,Suse9.0-9.3orDebian),HighRAMrequirements(particularlyiftensofsubjectsareusedinastudy),itislikelythatthe,computerwillneedtobe64bit.Thecomputershouldhaveatleasta1GHzCPUclock,1GBRAM,5GBswapand20GBfreeharddiskspace.,Ifmultiplenetworkedcomputers(oracomputercluster)areavailable,theregistrationstepscanbeparallelized,greatlyreducingthetotalcomputationtime.,Page40,白质纤维束的定量分析,FA:LeftCingulum(Red)RightCingulum(Blue),Parameterizationprocess,Gongetal,HumanBrainMapping,2005,Page41,同正常人相比，早期盲人的视放射白质扩散异常，FA值显著,降低，ADC和23显著升高。,早期盲人大脑白质扩散异常研究,Shuetal,HumanBrainMapping,2009,Page42,单张量模型的假设无法解决纤维交叉问题,纤维跟踪技术的准确性缺乏严格的评价体系,扩散张量成像的局限性,Page43,内容提纲,DTI的研究内容,DTI数据处理流程,DTIStudio,FSL:FMRIBsDiffusionToolbox,Page44,数据处理的基本流程,Page45,DWIfromScannerS0S1S2S3S4S5,S6,Page46,Preprocessing,DICOMdataconversion,Imagequalitycheck,Eddycurrentcorrection,Page47,内容提纲,DTI的研究内容,DTI数据处理流程,DTIStudio,FSL:FMRIBsDiffusionToolbox,Page48,DTIStudio,/Download&InstallUserManualMailinglist,Page49,LaunchingtheProgramandHardwareRequirement,DtiStudio-latest-x86.exeforWindowssystem,Morethan1GBRAMisrecommended,Page50,MainFunctions,ImageViewer,DiffusionTensorCalculationFiberTrackingandEditing3DVisualization,ImageFileManagement,ROIDrawingandStatistics,Page51,Howtodotensorcalculation,andfibertracking?,Page52,E:workTrainingExampleData,Rawdata:MRIcroNdcm2nii.exe,(.img,.hdr,.bvec,.bval),Eddycurrentcorrection:AIR,Tensor,FA,MDcalculation:DTIstudioFibertractography:DTIstudioROIselection,Page53,第一步：对原始DICOM数据进行格式转换。利用MRIcroN软件中的dcm2nii.exe工具，将DTI原始数据文件夹拖入，即可得到DTI扫描的梯度编码文件.bval和.bvec，以及转换后的NIFTI格式的图像文件（OutputFormat选择4DNIfTIhdr/img）。,Page54,第二步：对DTI图像进行头动和涡流校正。打开DTIstudio,File-MRIView3D,选中上一步得到的4D.img文件，ImageParameters中选择ImageFileFormat为Analyze，点击OK，然后在Image面板ImageProcessing区域选择AutomaticImageRegistration(AIR)，按图3进行设置，然后点击OK，等图像配准完成后，在Image面板的OrthogonalViews区域的文件下拉框中看到Air_开头的一系列文件，为校正后的DTI图像文件，点击Save，将Air_开头的所有文件选中，选择RawData，保存为一个4D的.dat文件。,MRIView,3D参数,Page55,AIR的参数设置,Page56,头动和涡流校正后的DTI图像保存,Page57,第三步：张量解算以及FA,ADC等扩散指标的计算。打开DTIstudio,File-DTIMapping,选择PhilipsREC格式，Continue，按图5进行参数设置，Addafile中选中上一步保存的4D.dat文件，点击OK，在,DtiMap面板的Calculation区域选中Tensor,ColorMapetc.（计算ADC值选择ADC-Map），根据图像选择噪声水平，点击OK，然后等DTIStudio算完后在Image面板的OrthogonalView区域可看到计算出来的各种扩散属性文件。对于想要保存的文件，如FA,EigenVector-0，ColorMap-0等可以分别进行Save（.dat格式），便于下一次查看和使用。,Page58,DtiMap面板进行张量解算,Page59,各种扩散属性的显示,Page60,第四步：纤维跟踪及可视化。第一种方法：基于前面步骤，在DtiMap面板的FiberTracking区域点击FiberTracking，然后进行参数设置，点击OK，就会进行基于全脑体素（FA0.2）的纤维重建；第二种方法：如果上一步已保存FA和EigenVector-0文件，可重新打开DTIStudio,File-Fiber-Tracking，选上FA-Map文件和PrincipleVector文件，并进行参数设置，点击OK，就会进行基于全脑体素（FA0.2）的纤维重建。通过任何一种方法，算完后右下角会出现Fiber面板，再此面板中可以对特定纤维束进行显示和编辑，并可以对纤维属性进行统计分析。,Page61,纤维跟踪方法2,Page62,纤维跟踪的参数设置,Page63,重建纤维束的可视化,Page64,Page65,Majorwhitemattertracts,Reference:WakanaS,CaprihanA,PanzenboeckMM,etal.,Reproducibilityofquantitativetractographymethodsappliedtocerebralwhitematter.Neuroimage,2007,36(3):630-644.,Page66,纤维属性的统计分析,Page67,NewModules,ROIEditor,ROIdrawing,selectionandmanip