DTI扩散张量的一种稳健估计方法

DTI扩散张量的一种稳健估计方法DTI Diffusion Tensor Estimation Using Robust MethodDiffusion Tensor Imaging (DTI) ha

DTI 扩散张量的一种稳健估计方法 DTIDiffusionTensorEstimationUsingRobustMethod DiffusionTensorImaging(DTI)hasbeenwidelyusedin medicalimagingtostudythemicrostructureoftissues, particularlyinthebrain.TheDTItechniquecanprovidevaluable informationonthediffusionbehaviorofwatermoleculesin biologicaltissues,indicatinghowtheyareaffectedbydifferent structures,suchasaxons,myelin,andcellmembranes.DTIis basedontheestimationofthediffusiontensor,whichsummarizes thediffusionpropertiesofwatermoleculesinagivenvoxelor regionofinterest. TheestimationofthediffusiontensorfromDTIdatacanbe challengingduetothepresenceofnoise,motionartifacts,and othersourcesofbias.Traditionalmethodssuchasthe least-squares(LS)ortheweightedleast-squares(WLS)method canprovidereliableresultsinidealconditions,wherethedatais noise-freeandthemotioniswell-controlled.However,in real-worldscenarios,thesemethodscanbehighlysensitiveto outliersanddataperturbations,leadingtoinaccuratetensor estimates. ToaddressthelimitationsoftraditionalDTImethods,several robustestimationmethodshavebeenproposedinrecentyears. Thesemethodsaimtominimizetheimpactofoutliersandnoise ontheestimatedtensorbyreducingtheirinfluenceonthecost functionusedtoestimatethetensor.Amongtherobustmethods, therank-deficientthresholding(RDT)andthereweightedL1-norm (RL1)methodshaveshownpromisingresultsinimprovingthe robustnessandaccuracyofDTItensorestimation. TheRDTmethodisbasedontheideaofthresholdingthe eigenvaluesofthediffusiontensor,removingsmalleigenvalues thatarelikelytocorrespondtonoisyanderroneoussignals.This methodappliesaregularizationparameterthatcontrolsthe degreeofthresholding,balancingthebias-variancetrade-off.The thresholdingoperationcanreducethedimensionalityofthe tensorspace,leadingtoefficientinversionanddecompositionof thetensor.TheRDTmethodhasbeenshowntoimprovethe accuracyandrobustnessofDTItensorestimation,particularlyin thepresenceofmotionandnoise.