论文部分内容阅读
PURPOSE VariousT1-weightedstructuralbrainimagingsequenceshavebeenoptimizedforquantitativemorphometricanalyses toaccuratelymeasurethevolumeofbrainstructuresthatcanbeaffectedbyvariousneurologicaldiseases.Themagnetizationprepared rapidgradientecho(MPRAGE)sequenceisoneofthemostcommonlyusedsequencesforthispurpose.Inorderto reachabettertissuecontrast,manyresearchersalsoexploredtheMagnetization-Prepared2RapidAcquisitionGradientEcho(MP2RAGE)andthemulti-echoMPRAGE(MEMPRAGE)sequences.Withthegrowingnumberofmorphometrytoolsthatcan “blindly”processanyT1-weighteddataset,itbecomesimportanttoassesstheeffectofsuchsequencesandusedhardwareon resultingvolumeestimates.Therefore,thegoalofthisworkistoexemplarilyanalyzethewithin-andbetween-sequence reproducibilityofbrainvolumesestimatedbytheMorphoBoxprototype. METHODS Whole-brainimageswereacquiredfrom12healthyvolunteers(agerange23-30yo,6female)at3T(MAGNETOM Skyra,SiemensHealthcare,Germany).Wechosetwotypesofcoils(20-channelhead/neckand32-channelhead),andcombined themwiththethreedifferentsequences,sothereweresixexperimentalconditionsintotal.Theorderofthesixexperimental conditionswasrandomizedperparticipant.Sequenceparameterswereasfollows: 1.MPRAGE:TI/TE/TR/flip/BW/echospacing=900ms/2.96ms/2300ms/9°/240Hz/7.1ms,FOV=256×256mm2. 2.MEMPRAGE:TI/TE1/TE2/TE3/TE4/TR/flip/BW/echospacing=900ms/1.69ms/3.3ms/4.91ms/6.52ms/2300ms/9°/890Hz/9ms, FOV=256×256mm2. 3.MP2RAGE:TI1/TI2/TE/TR/flip1/flip2/BW/echospacing=700ms/2500ms/2.9ms/5000ms/4°/5°/240Hz/6.9ms,FOV=256×256 mm2. Aback-to-backrepeatscanwasperformedforeachsequence.Quantitativevolumeestimatesforwhole-brain,subcorticaland corticalregionswereobtainedfromtheMorphoBoxprototypeusingdefaultparametersexceptforMP2RAGEwhereauxiliary imagewasprovidedforskull-strippingpurposes.Thevolumevaluesfromeachbrainregionwereusedtocalculatethetest-retest variability(TRV)(TRV=|M1-M2|*2*100/(M1+M2)),wherelowTRVvaluesmeanhighreproducibility.AFriedmantestwasusedto comparetheTRVamongthesixcoil-sequencecombinations.TheabsolutevolumedifferencesweremeasuredusingStudent– Newman–Keuls(SNK)analysis.Contrast-to-noiseratio(CNR)wasadditionallyassessedonalldatasetsusingthesegmentation mapsprovidedbyMorphoBoxas:CNR=(μGM- μWM)2/(σGM 2+ σWM 2) where μ and σ2 are mean intensity and variance within tissues gray matter (GM) and white matter (WM). Between-sequences CNR differences were assessed using a t-test with Bonferroni correction for multiple comparisons. All statistical analyses were performed using SAS 9.4. RESULTS Comparison of TRV values for whole-brain volumes shows that there is no statistical difference among the six combinations (p>0.05). Most TRV values are less than 3% with small standard deviation. The MP2RAGE and MEMPRAGE used with the 20-channelhead-neckcoilhaveslightlybetterreproducibilitythantheothercombinationsifweconsiderthemeanvalue andstandarddeviationofeachgroup.Theanalysisofthesubcorticalregionsshowssimilarresults.Wealsosummarizedthe regionswithmaximumandminimumTRVaveragesandlistedthethreeregionswithhighestvaluesandthreeregionswithlowest valueswitheachcoil-sequence.Thesegmentationandmeasurerepeatabilitydidntshowmuchdifferenceforeverybrain regions.Comparingtheabsolutevolumesobtainedbythedifferentsequences,theMP2RAGEshowssignificantlylargerGM volume(p<10-4)andsignificantsmallerWMvolume(p<10-4)thanbothMEMPRAGEandMPRAGE.TheGM-WMCNRof MP2RAGEwassignificantlyhigher(p<10-15)thantotheothertwo. DISCUSSION Theresultsshowthatforquantitativeanalysisofbrainstructures,thedifferentcoils(20channels,32channels)and sequences(MPRAGE,MEMPRAGE,MP2RAGE)didnothavealargeimpactonthereproducibilityofthevolumeestimates obtainedbytheMorphoBoxprototype.Theseoutcomescanhelptoguidethecoilandsequencechoicesformorphologyanalysis andautomaticvolumemeasuring.Cross-modalitycomparisonsshowthatMP2RAGE-basedGMandWMvolumesarerespectively largerandsmallerthaninbothMEMPRAGEandMPRAGE.AplausibleexplanationmaybethesignificantlyhigherGM-WMCNR inMP2RAGE.Furtherresearchwillaimtoinvestigatecorrectionstrategiestocompensateforvolumebiaseswhenpooling imagingdatafromdifferentsequences,e.g.forlongitudinalanalysisormulti-centricstudies.