《Multi-modality liver image registration based on multilevel B-splines free-form deformation and L-BFGS optimal algorithm.》.pdf

J．Cent．SouthUniv．f2014121：287—292 DOI：10．1007／sl1771—014—1939一Y 垒Springer Multi—modalityliverimageregistrationbasedon multilevelB--splinesfree--form deformationandL--BFGSoptimalalgorithm SONGHong(宋红)，LIJia-jia(李佳佳)，WANGShu．1iang(t}N良)，MAJing—ting(马婧婷) 1．SchoolofSoftware，BeijingInstituteofTechnology，Beijing100081，China； 2．SchoolofComputerScience，BeijingInstituteofTechnology，Beijing100081，China ◎CentralSouthUniversityPressandSpringer-VerlagBerlinHeidelberg2014 Abstract：A new coarse—to—finestrategywasproposedfornonrigidregistrationofcomputedtomography(CT)andmagnetic resonance(MR)imagesofa1iver．Thishierarchicalrfameworkconsistedofanaffine仃ansformationandaB．splinesrfee—ofrm deofrmationfFFD)．TheaffinetransformationperformedaroughregistrationtargetingthemismatchbetweentheCTandMRimages TheB—splinesFFD transofrmation perfomr eda finerregistrationby correcting loca1motion deformation．In theregistration algorithm，thenormalizedmutualinformation(NMI)wasusedassimilaritymeasure．andthelimitedmemoryBroyden-Fletcher- Goldfarb—Shannon(L—BFGS)optimizationmethodwasappliedofroptimizationprocess．Thealgorithm wasappliedtothefully automatedregistrationofliverCTandMRimagesinthreesubjects．Theresultsdemonstratethattheproposedmethodnotonly significantlyimprovestheregistration accuracybutalsoreducestherunning time，which iseffectiveandefficientofrnonrigid — registration． Keywords：multi—modalimageregistration；affinetransformation；B—splinesfree-fomr deformationfFFD)；L—BFGS feautresisimportantand itisachallenging task．The lIntr0ducti0n