ERROR LOADING HTML FROM SOURCE (http://ncf.sobek.ufl.edu//design/skins/UFDC/html/header_item.html)

NEUROIMAGING OF MILD TRAUMATIC BRAIN INJURY

Permanent Link: http://ncf.sobek.ufl.edu/NCFE004849/00001

Material Information

Title: NEUROIMAGING OF MILD TRAUMATIC BRAIN INJURY A VOXEL-BASED ANALYSIS OF SMRI AND DTI DATA
Physical Description: Book
Language: English
Creator: Robart, Graham
Publisher: New College of Florida
Place of Publication: Sarasota, Fla.
Creation Date: 2013
Publication Date: 2013

Subjects

Subjects / Keywords: Neuroimaging
Traumatic Brain Injury
DTI
MRI
Genre: bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Mild traumatic brain injury (mTBI) is a silent but costly epidemic, affecting millions of people worldwide every year. The inability for traditional medical imaging technologies to detect some types of damage that occur in mTBI has prevented the injury from receiving appropriate attention in research and clinical settings for decades. A continually growing body of research establishes diffusion tensor imaging (DTI) as an imaging modality sensitive to the subtle sequalae of diffuse axonal injury (DAI), common in mTBI. With the development of better imaging and analysis techniques for DTI, and the establishment of certain imaging biomarkers for identifying mTBI, a shift is occuring in the medical standards and practices for how we recognize and treat mTBI. Using T1-weighted MRI (sMRI) and DTI data obtained from the Mind Research Network (MRN), voxel-based analysis is performed to determine the relative merits of these data types in identifying mTBI and producing predictive biomarkers. These results demonstrate that diffusion tensor imaging (DTI) may be used to discriminate mTBI pathophysiology more effectively than standard imaging modalities such as T1-weighted magnetic resonance imaging (sMRI). A case is made for adopting DTI as part of standard clinical protocol in the diagnostic evaluations of head injury.
Statement of Responsibility: by Graham Robart
Thesis: Thesis (B.A.) -- New College of Florida, 2013
Supplements: Accompanying materials: CD containing results and materials DVD.
Electronic Access: RESTRICTED TO NCF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE
Bibliography: Includes bibliographical references.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The New College of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Local: Faculty Sponsor: McDonald, Patrick

Record Information

Source Institution: New College of Florida
Holding Location: New College of Florida
Rights Management: Applicable rights reserved.
Classification: local - S.T. 2013 R62
System ID: NCFE004849:00001

Permanent Link: http://ncf.sobek.ufl.edu/NCFE004849/00001

Material Information

Title: NEUROIMAGING OF MILD TRAUMATIC BRAIN INJURY A VOXEL-BASED ANALYSIS OF SMRI AND DTI DATA
Physical Description: Book
Language: English
Creator: Robart, Graham
Publisher: New College of Florida
Place of Publication: Sarasota, Fla.
Creation Date: 2013
Publication Date: 2013

Subjects

Subjects / Keywords: Neuroimaging
Traumatic Brain Injury
DTI
MRI
Genre: bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Mild traumatic brain injury (mTBI) is a silent but costly epidemic, affecting millions of people worldwide every year. The inability for traditional medical imaging technologies to detect some types of damage that occur in mTBI has prevented the injury from receiving appropriate attention in research and clinical settings for decades. A continually growing body of research establishes diffusion tensor imaging (DTI) as an imaging modality sensitive to the subtle sequalae of diffuse axonal injury (DAI), common in mTBI. With the development of better imaging and analysis techniques for DTI, and the establishment of certain imaging biomarkers for identifying mTBI, a shift is occuring in the medical standards and practices for how we recognize and treat mTBI. Using T1-weighted MRI (sMRI) and DTI data obtained from the Mind Research Network (MRN), voxel-based analysis is performed to determine the relative merits of these data types in identifying mTBI and producing predictive biomarkers. These results demonstrate that diffusion tensor imaging (DTI) may be used to discriminate mTBI pathophysiology more effectively than standard imaging modalities such as T1-weighted magnetic resonance imaging (sMRI). A case is made for adopting DTI as part of standard clinical protocol in the diagnostic evaluations of head injury.
Statement of Responsibility: by Graham Robart
Thesis: Thesis (B.A.) -- New College of Florida, 2013
Supplements: Accompanying materials: CD containing results and materials DVD.
Electronic Access: RESTRICTED TO NCF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE
Bibliography: Includes bibliographical references.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The New College of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Local: Faculty Sponsor: McDonald, Patrick

Record Information

Source Institution: New College of Florida
Holding Location: New College of Florida
Rights Management: Applicable rights reserved.
Classification: local - S.T. 2013 R62
System ID: NCFE004849:00001


This item is only available as the following downloads:


Full Text

PAGE 1

NEUROIMAGINGOFMILDTRAUMATICBRAININJURY:A VOXEL-BASEDANALYSISOFSMRIANDDTIDATA BY GRAHAMTANNERROBART AThesis SubmittedtotheDivisionofNaturalSciences NewCollegeofFlorida Inpartialfulllmentoftherequirementsforthedegree BachelorofArts UnderthesponsorshipofDr.PatrickMcDonald Sarasota,Florida May,2013

PAGE 2

Acknowledgements IamgratefultotheMindResearchNetwork,VinceCalhoun,AndyMayer,andJosefLing forprovidingmewithaccesstodata,softwaretoanalyzethedata,andguidanceinconducting neuroimagingresearchexperimentdesignandanalysis.WithouttheinternshipatMRN,thisthesis wouldneverhavebeenpossible.IwouldalsoliketoacknowledgeDr.PatrickMcDonaldforhis guidanceandmentorshipduringmyundergraduateexperienceatNewCollege.Iwouldliketo thankmybaccalaureatecommitteemembers,Dr.GordonBauerandDr.AlfredBeulig,andallthe professorsIhavehadtheprivelegeoflearningfrom.Thanksgotoallmypeers,whosharedinthe processandexperiencethroughtheirownworks.Finally,Iwouldliketothankmyfamily,Taylor Robart,RutiRobart,CaroleCalderone,PeggiandFrankIebba,andeveryonewhohashelpedme reachwhereIamtoday.Itisbecauseofallofyourguidance,patience,andlovethatIbecamewho Iam,andforthatIwillbeenternallygrateful. ii

PAGE 3

Contents 1Introduction1 IAReviewofTBIPathology,NeuroimagingModalities,Methodology,andResearch 7 2StructuralPathologiesofTraumaticBrainInjury8 3SummaryofVariousNeuroimagingModalitiesinmTBIResearch18 3.1ImagingModalitiesforDetectionofmTBIPathology................18 4T 1 -WeightedMagneticResonanceImagingsMRI23 4.1DevelopmentandMechanics.............................23 4.2MathematicsandMeasures..............................25 5DiffusionTensorImagingDTI26 5.1DevelopmentandMechanics.............................26 5.2MathematicsandMeasures..............................27 5.3DTIdetectionofmTBI................................29 6StatisticalAnalysisTechniquesandSoftware32 6.1AlgorithmsandStatisticalMethods..........................32 6.1.1StatisticalParametricMapping.......................32 6.1.2Voxel-BasedMorphometryAnalysis.....................33 6.1.3IndependentComponentAnalysisICA...................34 6.2StatisticalAnalysisSoftware.............................35 6.2.1AnalysisofFunctionalNeuroImagingAFNI................35 6.2.2StatisticalParametricMappingSPM8Program..............36 7DataAcquisitionandPreprocessing36 7.1Equipment......................................37 7.2FileFormats......................................38 7.2.1DICOMDigitalImagingandCommunicationsinMedicineFiles.....38 7.2.2NeuroimagingInformaticsTechnologyInitiativeNIfTIFiles.......39 7.2.3HEADandBRIKFiles............................39 7.3Noise:Sources,Prevention,andCorrection.....................39 7.3.1Thermalnoise.................................40 7.3.2Physiologicalnoise..............................40 7.3.3Systemnoise.................................42 7.4Normalization,Smoothing,andSegmentation....................42 7.4.1SpatialCoregistrationNormalization...................43 7.4.2MatterSegmentationGrey,White,CSF..................44 7.4.3SpatialFilteringSmoothing........................45 IIAStudyofStructuralMRandDiffusionTensorImagingofmTBI49 iii

PAGE 4

8MethodsandAnalyses49 8.1StructuralMRI....................................49 8.1.1ParticipantSelectionandDataAcquisition.................49 8.1.2ParticipantInclusionCriteriaandQualityAssurance............50 8.1.3ImageProcessingandStatisticalAnalyses..................51 8.2DiffusionTensorImaging...............................53 8.2.1ParticipantSelectionandDataAcquisition.................53 8.2.2ParticipantInclusionCriteriaandQualityAssurance............54 8.2.3ImageProcessingandStatisticalAnalysis..................54 9Results 56 9.1StructuralMRIResults................................56 9.1.1NeuropsychologicalandClinicalMeasures:................56 9.1.2DemographicCorrelations:.........................57 9.1.3SignicantClusterRegionsofInterest:...................59 9.2DiffusionTensorImagingResults..........................61 9.2.1NeuropsychologicalandClinicalMeasures:................61 9.2.2DemographicCorrelations:.........................61 9.2.3SignicantFractionalAnisotropyClusterRegionsofInterest:.......61 9.2.4SignicantMeanDiffusivityClusterRegionsofInterest:..........63 10Discussion66 11Conclusion70 iv

PAGE 5

NEUROIMAGINGOFMILDTRAUMATICBRAININJURY: AVOXEL-BASEDANALYSISOFSMRIANDDTIDATA GrahamTannerRobart NewCollegeofFlorida,2013 ABSTRACT MildtraumaticbraininjurymTBIisasilentbutcostlyepidemic,affectingmillionsofpeople worldwideeveryyear.Theinabilityfortraditionalmedicalimagingtechnologiestodetectsome typesofdamagethatoccurinmTBIhaspreventedtheinjuryfromreceivingappropriateattentionin researchandclinicalsettingsfordecades.AcontinuallygrowingbodyofresearchestablishesdiffusiontensorimagingDTIasanimagingmodalitysensitivetothesubtlesequalaeofdiffuseaxonal injuryDAI,commoninmTBI.Withthedevelopmentofbetterimagingandanalysistechniquesfor DTI,andtheestablishmentofcertainimagingbiomarkersforidentifyingmTBI,ashiftisoccuring inthemedicalstandardsandpracticesforhowwerecognizeandtreatmTBI.UsingT 1 -weighted MRIsMRIandDTIdataobtainedfromtheMindResearchNetworkMRN,voxel-basedanalysis isperformedtodeterminetherelativemeritsofthesedatatypesinidentifyingmTBIandproducingpredictivebiomarkers.TheseresultsdemonstratethatdiffusiontensorimagingDTImaybe usedtodiscriminatemTBIpathophysiologymoreeffectivelythanstandardimagingmodalitiessuch asT 1 -weightedmagneticresonanceimagingsMRI.AcaseismadeforadoptingDTIaspartof standardclinicalprotocolinthediagnosticevaluationsofheadinjury. Dr.PatrickMcDonald DivisionofNaturalSciences v

PAGE 6

1Introduction Thehumanbrainhasbeenstudyingitselfsincetheearliestaccountsinrecordedhistory,andmany ofthesignicantinsightsintotheimportanceandfunctionofthebrainhavebeenmadebecause ofinjuriesanddisease.Ancientskeletonsandpreservedhumanremainspredatinganyhistorical recordshaveshownevidenceoftrepanation,theprocessbywhichaholeoropeningismadein theskull,exposingtheduramater. [9] Itisperhapstheoldestknownexampleofsurgicalprocedure supportedbyarchaeologicalevidence,andinsomeareasthepracticeoftrepanationmayhavebeen quitewidespreadandcommon.Forexample,prehistoricskullsfoundatahumanburialsitein Francedatingbackto6500BCE [8] ,exhibitsomeevidenceoftrepanation.Ofthe120humanremains uncoveredatthesite,40skullshadmarkingsoropeningswhosestructureappearedintentionaland notassociatedwithdamageortrauma.Someoftheskullsalsohadplatesorfragmentsofbonefused backintoplace,whichsuggeststhatsubjectssurvivedandrecoveredfollowingthetrepanation. Whileitisimpossibletoknowwhatbeliefssurroundedthepracticeoftrepanationinprehistoric contexts,itindicatessomerecognitionofthebrainasthesourceofcertainpathologies,andofearly attemptsattreatment. 1 ItwasnotuntilthewritingofTheEdwinSmithPapyrusthatanywrittenevidenceexisted documentingthebeliefthatthebrainhadsomeeffectorroleincognitionandcoordination.The existingdocumentisdated1500BCE,butmaybeacopyofanearliertextwritten1000yearsbefore 1 Theancientpracticeoftrepanationisthoughttohavebeenacrudemethodtotreatsometypesoftraumaticbrain injuries,andepilepticseizures.Itisaprocedurethatremainshighlyeffectiveeventodayinrelievingpressureinhematoma andintracranialswelling. [9] 1

PAGE 7

that [8,9] .TheEdwinSmithPapyrusalsocontainstheearliestknownreferencetothebraininwritten history.ItdetailssomemedicinalpracticesoftheancientEgyptianscirca3000-2500BCE [12] Amazingly,theEgyptiansanatomicalandphysiologicalknowledgeaswellastheirmethodsof clinicaltreatments,matchedorsurpassedthatofHippocrates'steachingsovertwothousandyears later.TheEdwinSmithPapyruscommunicatesanunderstandingoftherelationbetweenheadinjury andtheonsetofsymptomssuchasparalysis,retrogradeamnesia,andanterogradeamnesia.Italso notesthelossofmotorfunctionandsensoryreceptionthatoccursinspinalinjuries,andhowthey correspondtorespectiveregionsofthebodybasedonwhichvertebraeweredamaged. [12,9] Theregardforthebrainastheseatofcognitionandconsciousnesshasbeenmade,lost,and rediscoverednumeroustimessincethewritingoftheEdwinSmithPapyrus.Infacttheideathat consciousnessresidesintheheartoreveninthe'soul',meaningnotanywhereinthephysicalbody butonaseparateplaneofexistence,waswidelyacceptedbyGreciansocietyfromtheteachingsof Aristotle.Aristotlebelievedthebrainwasoflittleimportance,describingitasanorganforcooling thebloodbasedonhisobservationthatasignicantportionofthebody'sbloodcirculatedthrough thebrain. [97] Aristotle'ssystematicdenialofthebrainasthecenterofsensationormovementcontinuedtoinuencepopularsocietalviewsformanycenturies.Thismaybeinpartduetoacultural aversionoftheancientGreekstotouchingordissectingthebrain,preventingfurtherinvestigation intoitsfunctions. Pythagorascirca550BC,knownforthePythagoreantheoremisregardedasoneoftherst toproposethatthoughtprocessesandthesoulwerepartiallylocatedinthebrain,andnotsolely theheart. [98] Inaddition,AlcmaeonofCrotonwasoneoftherstwriterstochampionthebrainas thesiteofsensationandcognition. [97] Despitetheseearlyassertionsofthebrainsrole,themajority ofhealersandphysiciansdidnotadoptthisideauntilafterHippocrates.Hippocrates,wholived around400BCE,isoftenregardedasthefatherofwesternmedicine.MuchofHippocratesinsights camefromhisextensiveworktreatingpatients,ratherthanonlyfromexistingmedicalteachings. OneofHippocratesmostsignicantcontributionstoneurosciencearosewhiledealingwithpatients sufferingheadinjuries;heobservedparalysisconsistentlymanifestedonthesideofthebodyoppositethatoftheheadinjury,leadinghimtodiscovercontra-lateralizationofmotorfunctionwithin thebrain.Hippocrateshimself,andmostmajormedicaldisciplinesdescendantfromHippocratic medicine,subscribedtothecephalocentrichypothesis,whichstatesthatthebrainisthesolesource 2

PAGE 8

ofallconsciousness,sensation,andmotorcontrol.Thiswasacriticalturningpointinhistoryfor thewayinwhichtheeldofmedicineapproachedthebrain. [9] Despitebeingequippedwiththisinsight,progressinthestudyofthebraincontinuedslowlyfor thenextmillennium.Mosthumanbrainresearchandinvestigationwasdoneviaautopsy,orinpreservedsamplesofthelong-deceased.Studyingthelivingbrainwasdifcultwithoutresortingtoinvasiveproceduressuchasneurosurgeryorablationofnervoustissues,whichwerepracticedmainly onanimals.LeonardodaVincihimselfiscreditedasbeingthersttopithafroginordertostudy itsphysiologyandthestructureofitsnervoussystemneartheendofthe15thcentury. [102] Leonardo alsomadewaxcastsofthebrainandventriclesofanox,respectively,tostudyandcompareto humanmodels.Hisinvestigationswerefaraheadoftheexistinganatomicalknowledge,andhis secrecykeptthemfromreallycontributingtoscienceofthetime.Additionally,onlysomuchof thelivingbrain'scomplexitycanbegleanedfromitsinanimateform.Forexample,manyofDa Vincisideasaboutthesensesandreasoningsuggestedtheventriclesastheimportantstructuresof theconsciousmind.Inlivinghumans,thepracticeoflobotomyorablationwasusuallyrestricted totheseverelymentallyill,asanattempttoalleviateavarietyofmaladies.Eventoday,thecrude methodofseveringthecorpuscallosumcanprovidesignicantreductionofsymptomsinthose sufferingfromsevereepilepsyandseizures,toadegreethathasyettobereproducedsolelyby pharmacologicalmeans. Overtime,thedenitivevalueofunderstandingthelivingbrainencouragedthepursuitofother non-invasivemeansofinvestigation.Itwaspartlyduetothispursuitanditsappealthatphrenologybecamesopopulararoundtheturnofthe19thcentury.Phrenologywasthepracticeofusing measurementsoftheheadandskulltomakeinferencesaboutthepersonalityandcharacteristicsof aperson,basedontheideathatitrevealedtraitsofthebrain.Theoretically,phrenologyallowed forthecollectionoflargequantitiesofdatawithoutinvasiveprocedures,inanattempttomake inferencesabouttheunderlyingbrainanditsfunctionalcharacteristics.Whenphrenologyhadrun itscourse,andwasconsequentlyrejectedasungroundedpseudoscience,thatpursuitreturnedto thestudyofdiseaseandinjuryastheprimarymeanstoexplainthefunctionofthebrain.Patients sufferingfromlesionsinthebrainduetodiseaseorinjuryoftenexhibitedchangesinbehavior thatimplicatedthelocalizationoffunctionwithinthecentralnervoussystemCNS.Forinstance, subjectssuchasPhineasGageandpatientHMprovidedinvaluableinsightintothebrainregions 3

PAGE 9

implicatedinmemory,emotion,andreasoning. [9] HM'sunexpectedlossofshort-termmemoryformationfromhisbilateralhippocampectomy,andGage'srailspikeaccidentareunfortunatelyonly casestudies.Theirexperienceswerealsobothveryinvasive,thedamagefromwhichisnotreadily reproducible,eithermorallyorinpractice.Assuch,itisdifculttomakeinferencesregardingthe generalpopulationfromsuchindividualcasestudies.StudyoflargercohortsofCNSdiseaseand disorderhavebeendoneforalongtime,butinthepast,large-scalestudiesofCNSinjuryhave beenpartiallyconfoundedbyvariabilityinthetimingandcausesofcivilianheadinjuries.Military forcesontheotherhand,moreregularlyobservelargenumbersofinjuries.Oftentheseinjuries aresustainedwithinsimilartime-frames,andmanyofthemmayexhibitsimilarpathologies.This makesthemidealforcomparisonandinvestigation. Throughouthistory,warhasbeenthecauseofavastportionofallinjuriessustained.Aswars havebeenconductedbyincreasinglylargerpopulations,andusingincreasinglydangerousweapons suchasexplosivesandguns,thenumberofwartimeinjurieshavealsoescalated.Thesemyriad traumawouldoftenprovefatal,especiallyinthecaseofheadinjuries.However,continuingprogress inmedicalpracticesalsomeantthatanever-growingportionofthosewhosufferedcombat-related injurieswouldsurvive.Wartimeheadinjuriesandtheirtreatmentshavethereforebeenthesource ofmuchmaterialinmedicalresearchontraumaticbraininjury.ForexampleduringWorldWarI, followingthedevelopmentofnewantiseptictechniquesinthelate19thcenturyandmoreeffectiveneurosurgicaltechniques,themortalityofaduralpenetrationheadwoundwasonly35%. [3] Priortothis,thelargemajorityofopenheadwoundsinpatientsprovedfatal.DuringtheCivil War,gunshotwoundstotheheadwereseeninlargenumbers,andalthoughaccuratestatisticsformortalityratesarenotavailablefortheCivilWar,survivalwasuncommondue tocomplicationandinfection.Withthewidespreaduseofhelmetsandothermeansofprotection fromheadinjuryinsubsequentwars,thenumberofmilder,non-fatalheadinjuriessawamarked increase. [9,7] Itwaswiththeadventofnon-invasivemedicalimagingaroundthistimethatserious investigationintobraininjurybegan.WiththeabilitytoobserveinvivopathologieswithCTscans, itwasquicklyrecognizedthatfastreliefofahematomawascriticaltoimprovingmortalityrates fromsevereheadinjury. [79] IntheKoreanWar-1953,mobilemedicalteamscapableofevacuatinghematomasweredeployedamongUSsquadrons,andthisresultedinadropintherateof injury-associatedmeningiocerebralinfectionsfrom42%tolessthan1%. [7] BythetimeoftheViet4

PAGE 10

namWar,itwasestimatedthat40%ofUScombatfatalitieswerebecauseofheadandneckinjuries, and14%ofthosesurvivinginjuryexhibitedlong-termimpairmentsandcomplicationswhichare symptomaticofwhatwouldlatercometobeknownastraumaticbraininjuryTBI. [5,32] Traumaticbraininjuryisdenedasdamagetothebrainresultingfromexternalmechanical force,suchasrapidaccelerationordeceleration,directimpact,blastwaves,orpenetrationbya projectile. [13,10] Brainfunctionistemporarilyorpermanentlyimpairedandcertainstructuraldamagemaynotalwaysbereliablydetectedusingmoderndiagnosticsandtechnology.Inordertobetter recognizeandtreatthediversesequalaeassociatedwithheadinjury,anorganizationwasestablished toconductresearchandtoimprovetreatmentofveteransandciviliansforTBI.Theformationof theDefenseandVeteransBrainInjuryCenterDVBICin1992wasacollaborationbetweencivilianorganizationsandtheDepartmentsofDefense,andVeteransAffairs. [7] Itsgoalwastoestablish clinicaldenitionsandstandardsofTBI,andprovideresourcestoimproveawarenessandrecovery.Originalclassicationsandguidelinesreliedheavilyonsymptomatologyandthepathologies detectablebyimagingtechnologiesavailableatthetime.Thesemeasuresweredevelopedwitha cutoffofseverityinmind,similartootherclinicaldiagnosessuchaspsychiatricevaluations.This wasmeanttoappropriatelylimittheoccurrenceofafalsediagnosisinanotherwisehealthypatient. ItwasbelievedthatTBIwascategoricallydifferentintermsofpathophysiologyfromconcussion, andshouldbedistinguishedaccordingly.However,furtherexperiencewithcombatrelatedTBIbegantosuggestthatmanysurvivorswithpathologybelowthecriteriathresholdsforTBIdiagnosis werestillexperiencingmeasurablecognitiveimpairments. Theprevalenceofexplosiveandblast-relatedinjuryandheadtraumaareallhigherfortheIraq andAfghanistanwarsthananywarpreviously. [53,32] Thisislikelyduetotheincreaseduseofimprovisedexplosivedevices,artillery,rocketpropelledgrenades,andminesinunexpectedorsudden instances.InfactitwasingreatpartbecauseofthenumberofIraqandAfghanistanveteransreturningwithoutdiagnosedTBIwhilestillsufferingfromtheimpactoftheirinjuriesthatthedenition forTBIwasexpandedtoincludemildTBI.mTBI [54] Itisestimatedthatuptoastaggering20%of currentlydeployedservicemembersmayhavemildTBIfromblastinjury,andhasthereforebeen termedthesignatureinjuryofthewar. [103,4,7] Concussiveinjuryiscurrentlybeingsubsumedinto thedenitionofmildTBI,whichmayimproveourunderstandingandtreatmentofsuchinjuries. Previously,concussiontreatmentguidelinesmostlyconsistedofgeneralprotectivemeasuressuch 5

PAGE 11

asmaintainingpatientconsciousnessduringtheacutephaseofinjury,andthenprescribedrest. ThemajorityofcasesofTBIinbothcivilianandcombatrelatedsettingsarecategorizedas mTBI,aclassicationbasedprimarilyonthecharacteristicsoftheacutesequelaefollowingtheinjury.TheworkingdenitionofwhatconstitutesmTBIremainsunderconstantrevision,yetpeople stillexhibitmTBIinsomedetectableways. [94,10,91] DozensofmeasuresandcriteriaformTBIhave beendevelopedbyvariousinstitutionsovertheyears,eachdifferingslightly.Thesetestsandmeasureshaveallhadvaryingstrengthsofcorrelationwiththeunderlyingpathologiesobserved,which hashelpedinformthemostpredictivecriteriainfuturedenitions.ThemostrecentofcialdenitionofmTBIisprovidedbytheDepartmentofDefense/DepartmentofVeteransAffairsDOD/VA ClinicalPracticeGuideline. [94] Theirdenitionhasadoptedthefollowingcriteria: 1.BriefminutesorlesslossofconsciousnessLOCfollowinginjury 2.Briefupto24hoursalterationofconsciousness-confusion,poorresponsiveness,poor motorcoordination 3.Post-traumaticamnesiaPTAlimitedtotherst24hoursfollowinginjury 4.GlasgowComaScorebestscoredwithintherst24hoursof13to15=normal 5.AnormalappearingbrainoncomputedtomographyCTscan Eachyear,morethan1.7millionpeopleintheUnitedStatesexperienceaTBI,witharound75% oftheseinjuriesbeingcategorizedasmTBI. [53] Thisnumberislikelyanunderestimatebecauseit doesnotincludethosewhoareseeninprivateclinicsorbyprimarycarephysicians,nordoesit includethosewhodonotseekanymedicaltreatment. [76] TheestimatedeconomiccostofTBIis alsoenormous,withmorethan60billiondollarsbeingspentannuallyintheUStreatingTBIsince 2000. [101,32] Thiscostmaynowbewellover75billionperyear,withanestimated44%ofthecost relatedtodiagnosedcasesofmTBIalone. [32,15] DespitethesesignicantcontributionsofmTBI totheoverallimpactofTBIonthepopulation,mostmodernneuroimagingdiagnosticequipment isunabletoconsistentlydetectmTBI [40,15] .Inaddition,somesubjectssufferingfrommTBImay 6

PAGE 12

remainasymptomaticinwaysthatpreventtheiridenticationevenusingtheDOD/VAclinicalpracticeguidelinesabove.AnotherimportantfactorconfoundingmTBIrecognitionanddiagnosisisthe highrateofco-morbiditywithposttraumaticstress-disorderPTSD. [5,4,13,103] Therefore,alarge portionofthepopulationafictedbymTBImayremainundiagnosed,andsubsequentlydonotreceiveappropriatetreatmentsfortheirinjury.Forthesereasons,itisimperativeforfuturediagnostics todetermineamoreobjectivedescriptionoftheunderlyingneurophysiologythatconstitutesmTBI. MoreaccurateandwidespreadrecognitionofmTBImayincreasetherateofindividualrecoveryby allowingmoreappropriatetreatmentstobeadministeredcasebycase.Suchimprovementscould signicantlydecreasetheeconomicburdenofTBIonoursociety,morethanjustifyingthecomparativelysmallcostofcontinuedresearch. Inthisthesis,westudytheuseofmodernneuroimagingtechniquesforidentifyingbiomarkersof mTBI.UsingdataobtainedfromtheMindResearchNetworkMRN,wedemonstratethatdiffusion tensorimagingDTImaybeusedtodiscriminatemTBIpathophysiologymoreeffectivelythan standardimagingmodalitiessuchasT 1 -weightedmagneticresonanceimagingsMRI.Analysis ofsMRIdataconrmssomepreviousndingsintheliterature,albeitwithalargerpopulationsize thanpreviouslyexamined.WefoundsMRItobeoflimitedutilityindiagnosingmTBI,sincesmall scalepathophysiologicaltraitssuchasdiffuseaxonalinjuryDAIandshearingdamagearedifcult todetectwiththisimagingmodality.AnalysisofDTIdata,ontheotherhand,hasbeenshownby previousliteraturetohaveasignicantpredictivepowerinidentifyingDAIandotherphysiological characteristicsofmTBI. [15,33,27,26,25,1] Theremainderofthisthesisisorganizedasfollows:First,thepathologiesoftraumaticbrain injury,thetechnologiesusedtoidentifyanddiagnoseit,andthestateofcurrentresearchwithinthe eldareexaminedandreviewed.Second,astudyisconductedthatperformsanalysisonneuroimagingdatausingcurrentmethodsofstatisticalparametricmappingandvoxel-basedcomparisons. 7

PAGE 13

PartI AReviewofTBIPathology,Neuroimaging Modalities,Methodology,andResearch 2StructuralPathologiesofTraumaticBrainInjury ImmediatelyfollowingTBI,thereareseveralmetabolic,hemodynamic,structural,andelectric changesthatalternormalcerebralfunction.Thesealterationscanincreasethebrain'svulnerability torepeatinjuryandlong-termdisability. [44] ClinicalclassicationsofstructuraldamageinTBIhave beenmadetobetterunderstandanddiagnosethevarioustypesofabnormalitiesthatoccur.TBIis usuallyclassiedintoprimaryandsecondaryinjuries.Primaryinjuriesarecommonlythedirect resultoftraumatothehead.Secondaryinjuriesariseascomplicationsoftheprimarylesions.Any combinationofthesepathologiesmayoccurco-morbidly,orinisolation,dependingonthenature ofthetraumasustained. Primaryinjury PrimaryinjuryinTBIhasnumerousimmediateeffectsincludingcontusion,hematoma,disruptionofthebrain'sstructuralintegrity,intercellularhemorrhagingleadingtothedevelopmentof hematoma,theconcentrationofbloodinalocalizedregion.Notallthesetypesofbleedingare immediatelythreatening,insomecasesbleedingcansustainforweeks,andeventuallyproducesecondaryinjuriessuchasbloodclotsandischemia.Primaryinjuriesarecategorizedbetweenintra andextra-axialinjurytypes. Extra-axialinjury Anyheadinjuryresultingindamageoutsideofthebraintissueisconsideredanextra-axialinjury. Thisincludesdamagetothesubarachnoid,subdural,epidural,andventricularspaces. [103] Epiduralhematomaiswhenbleedingoccursinthespacebetweentheduramaterandtheinterior 8

PAGE 14

oftheskull.Thebleedingoftenoccursfromrupturedarteriesinthetemporalmeningesdueto trauma.Themajorityofepiduralhematomaoccurwithinthetemporalregionipsilateraltothe siteoftrauma,highlycorrelatedwithimpactonthesideofthehead.Duetothehighpressure ofarterialveins,thebleedingandonsetofepiduralhematomacanbequiterapid.Peaksizeof epiduralhematomausuallyoccurswithin7-9hoursofinjury. [104] Theinnerskullwillnotyieldto theincreasedpressurecausedbytheadditionalvolume.Thesutureconnectionsoftheduramaterto theskullpreventdispersalofthebloodalongsidetheintercranialspace,andsothepocketofblood willinsteadexertinwardforceonthebrain.Thisresultsinaconvexor'lentiform'displacementof brainmatter,potentiallycrushinganddamagingtissues. Figure1:ACTscanofapatientsufferingfromalargeepiduralhematomalightgrey.Noticethe lentiformdisplacementofthedarkerbraintissues. Epiduralhematomamaypresentnosymptomsimmediatelyfollowinginjury,butasthepressure grows,impairedcognitionorunconsciousnesscanfollow.Trepanationandsurgicalremovalof thebloodtorelievepressurepreventsfurtherdamage,andgreatlyimprovesrecovery.Epidural hematomahavea15-20%riskoffatality. [104] Subduralhematomaoccurduetohemorrhageofsmallerbridgingveinswithinthemeningeal layer,betweentheouterduramaterandtheinnerarachnoidmater.Subduralhemorrhagingis muchmorecommonthanepiduralhemorrhaging,andoccursinmanycasesofmildTBIandnonconcussiveheadbumps. [90] Thesourceofbleedinginsubduralhematomaisusuallynotarterial,and 9

PAGE 15

solowerpressurebleedingmayoccuroveramuchmoreprolongedperiodoftimethaninepiduralhematoma.Thespeedofonsetdetermineswhetherepiduralhematomaisconsideredacuteor chronic.Inacutecases,symptomsmaymanifestimmediately,orwithinminutesofinitialtrauma. Acutesubduralhematomaisoftenassociatedwithrapidacceleration/decelerationshearingandoccursco-morbidlywithotheraspectsofsevereTBI. [90] Themortalityrateassociatedwithacutesubduralhematomaisaround60to80%,makingitoneofthedeadliestformsofinjuryassociatedwith TBI. [13,53] Chronicsubduralhematomaontheotherhand,presentsoverthecourseofdaysorweeks andmaynotbedetecteduntilmonthsoryearsaftertheinitialtrauma. [104] Withsuchslowbleeding, hemorrhagingmaystoponitsownshortlyafterinitialtrauma.Chronicsubduralhemorrhagingcan developasubduralhematomaoverthecourseofmultiplesmallbleedsinthesameregion,where tissuesarealreadypronetore-injury.Comparedtoacutecases,chronicsubduralhematomaoffer abetteropportunityforstoppingdevelopmentbeforedamageoccurs,andtreatmentsresultingood orcompleterecoveryincloseto78%ofcases. [53] Figure2:ACTscanofapatientsufferingfromasubduralhematoma,indicatedbyarrows.Notice theconcavedevelopmentalongtheskull. [90] Subarachnoidhemorrhagedescribesbleedinginthespacebetweenthesubarachnoidmembrane andthepiamater.Thisisusuallycausedbythespontaneousrupturingofamajorbrainarterydueto 10

PAGE 16

weakarterialwalls,orlesscanbeinducedbytrauma.Inthecaseoftrauma,subarachnoidhemorrhagingoftenoccursnearafractureoftheskullorcerebralcontusion.Subarachnoidhemorrhaging isaveryserioustypeofinjury,withsymptomsofsevereheadacheandothercognitiveimpairment developingrapidly.Themortalityrateisveryhigh,withcloseto50%ofcasesbeingfatal. [105] Risk ishighbecauserupturesinregionsbelowtheCircleofWillismayresultinischemia.Treatmentfor subarachnoidhemorrhageisthereforeverytime-sensitive,andthosewhodelayinreceivingmedical attentionareassociatedwithworseprognosis.Inmildcasesoftraumaticsubarachnoidhemorrhaging,symptomsmaybemistakenwithpainfromthetraumaitself,andmaygoundiagnosedfor longer. [105] Intraventricularhemorrhageoccurswhenbleedingleaksintotheventricularspacesofthebrain. Severityoftheinjuryisbasedonhowmuchtheventriclesarelledwithblood.Iftheventricles arecompletelylled,internalpressurebuildsandbeginstocauseseriousdamage.Thistypeof hemorrhagingoccursin35%ofmoderatetoseverecasesofTBI. [108] Luckily,headtraumamustbe somewhatseveretocauseintraventricularhemorrhage,andsoitdoesnotreallyoccurincasesof mTBI. Intra-axialinjury AnyheadinjuryresultingindamagetotissueswithintheCNSareconsideredintra-axialinjuries.Thisincludesaxonalinjuries,contusions,intracerebralhematoma,andarterialdissection. AxonalinjuryisthemostcommontypeofinjuryinTBI,anddescribesanytypeofphysical damagetotheelongatedaxonalbodiesofneuronsintheCNS. [8] Damagecanbesevereimmediate, causingtearingandseparationoftheaxonfromtheneuroncellbody,orcanoccurslowlyasaresult ofdegradationofcytoskeletalarchitecture.Shearingforcethataltersaxoplasmicmembranescanbe fromstretching,rotationaltorsion,orbending.Normallyaxonsarerelativelyelastic,buttheymust dososlowlyinordertomaintainstructuralintegrity.Rapidchangeaxoplasmicmembraneshape causesthecytoskeletontobecomebrittleordeformed,whichsubsequentlyimpairsaxoplasmic transportofproteins,neurotransmitters,andotherorganicmolecules. [44] Thedisruptionoftransport functionsresultsindelayedonsetofdamageandoftenleadstoapoptosisinneurons. 11

PAGE 17

Figure3:Anexampleofrotationalshearingresultinginimpairedaxonaltransportandeventual apoptosis. [110] ThemechanismfordiffuseaxonalinjuryDAIisshear-straindeformation,achangeinshapeof thebrainwithoutaresultantchangeinvolume.PathologystudieshaveshownthatDAIischaracterizedbymulti-focallesionsofaxonaldamage,scatteredthroughoutwhitematter,thatareoccasionallyassociatedwithhemorrhage. [87,49] Experimentalstudieshavealsodemonstratedthatdiffuse whitematterinjuriesmayoccurwithtraumaticbraininjuryirrespectiveofskulldeformation. [43,57] DAIoccursinupto48%ofpatientswithclosedheadinjuries [91,31] ,andiscausedprimarilyby theshearingforcesgeneratedduringrapiddecelerationoraccelerationofthehead,andmaynot eveninvolveanyimpacttrauma.Forexample,DAIoftenoccursinmotorvehicleaccidents,even whennoheadimpacttakesplace. [31] DAIpatternofexpressionisheterogeneousandbilateral,and frequentlyinvolvesthelobarwhitematteratthegray-whitematterinterfaceofthecerebralhemispheres.DAIhasbeenobservedtoregularlyaffectwhitematterinareasincludingthebrainstem, thecorpuscallosum,andsuperiorpeduncle.TheheterogeneousexpressionofDAImakesithard todetectreliably,andmaymanifestindifferentregionsonapatientbypatientbasis.Additionally,damagemayoccurinnumerousareas,eachexhibitingonlyminorstructuralchangeshence thenamediffuse.AlthoughDAIisrarelyfatal,itcanresultinsignicantneurologicalimpairment.DAIissocommoninTBIthatinthemajorityofmTBIcases,itmaybetheonlydetectable pathology. [87] Unfortunately,inmanymTBIcasesDAIisstillverydifculttodetectwithCTscan andsMRI. [89] Corticalcontusionsareessentiallybruisingofthebraintissues.Thesebruisesformfromsmall 12

PAGE 18

bloodvesselleaksofveinsandarteriescoveringtheparenchymaltissue,orfrommicro-hemorrhages resultingfromtrauma. [104] Whenthebloodvesselistorn,bloodescapesfromthevesselatarate thatisfasterthanthebloodthatcanbeabsorbedbythebrain.Consequently,corticalcontusions oftenresultinedemaandincreasedintracranialpressure.Theinjurycancauseadeclineinmental functioninthelongterm,andinseverecasesmayleadtobrainherniation.Thecontusionmaycause swellingofthesurroundingbraintissue,whichmaybeirritatedbytoxinsreleasedinthecontusion, Thecytotoxicedemausuallypeaksfourtosixdaysaftertheinitialtrauma. [66] Intracerebralhematomaiscausedbyasuddenleakageofbloodintocorticalbraintissues.The hemorrhagingoccursfromverysmallarteriesandpushesintothesurroundingbraintissue,causing injurytotheregionandsymptomsofastroke.Becausethebloodhasnospacetogoinsidethecorticaltissueslikeitdoesinextra-axialhematoma,displacementoftissuesandsubsequentdamage beginsalmostimmediately. [105] Intracerebralbleedingisalsoknownasintraparenchymalhemorrhaging.Sometimesbloodwillndaroutetotheventricles,andcauseintraventricularhemorrhaging.Symptomsusuallyinvolvesuddenonsetofstroke.Strokesymptomsincludesuddenweakness, blindness,lossofspeechandnumbnessonhalfofthebody. [10] Othersymptomsalsocaninclude seizureandlossofconsciousness.Theriskofdeathfromanintraparenchymalbleedintraumatic braininjuryisespeciallylowwhentheinjuryoccursinthebrainstem. [75] Intraparenchymalbleeds withinthemedullaoblongata,ontheotherhand,arealmostalwaysfatalbecausetheycausedamage tothevagusnerve,whichplaysanimportantroleinbloodcirculationandrespiration. [53,104] Arterialdissectionreferstoatearingandseparationoftheinnermostwallofanartery.Inintraaxialinjury,dissectionfrequentlyoccursinmajorcerebralarteriessuchasthecarotid,ornearthe CircleofWillis. [109] Asthetearbecomeslarger,itformsasmallblister-likeseparationcalleda falselumen.Bloodaccumulatesinsidethisblisterandcanleadtostrokeinsomecases.Blood canpoolinsidethewallofthearteryuntilitbeginstoimpedenormalbloodow,whichiswhen thepocketbeginstobereferredtoasapseudoaneurysm.Pseudoaneurysmscanleadtosymptoms ofstrokebypressingonbrainstructureslocatednearby,orbyrupturingandcausingmajorhemorrhagicstroke. [104] Whenrupturingoccurs,thepseudoaneurysmisoftenreferredtoasadissecting pseudoaneurysm.Alternatively,bloodinsidethefalselumencanclotandeventuallyprotrudeinto theareawherebloodnormallyowscausingthrombosis,whichleadstomajorischemiaofaffected brainregions.Patientsexperiencingthrombosisarealsoatriskforembolismtooccur,inwhich 13

PAGE 19

smallpiecesfromthegrowingbloodclotcanbreakoff,owupstream,andbecometrappedinside asmallerarteryinthebrain. [104] SecondaryInjury Damageandcomplicationscandevelopinbraintissuesafterthetraumaofprimaryinjury,leadingtofurtherissues.SecondaryinjuriesofTBIincludeanytypeofdysregulationofhealthybrain functionstemmingfromaninitialinjury.Secondaryinjuriescaneitherbeacuteorchronic,dependingontherelativespeedoftheironset.Secondaryinjuriestendtobeeasiertopreventandtreat withappropriatecarethananyprimaryinjurysequalae,butifcomplicationsdoarise,manyofthese conditionscanbedebilitating. Acute Anysecondaryinjurythatmanifestduringthe24hoursfollowingheadtraumaareconsidered acute.Thisincludesdiffusecerebralswelling,alsoknownascytotoxicedema,brainherniation, infarction,andinfection. Diffusecerebralswellingdisplaysapatternofheterogeneousexpressionsimilartothatseen indiffuseaxonalinjury.Exceptthatinthiscase,traumadoesnotleadtoshearingofneuroncytoskeletalstructure,butinsteadleadstoamaladaptiveuptakeofwatercontentbybraintissues.The twomaintypesofedemathatleadstocellularswellingarevasogenicandcytotoxicedema. [114] In vasogenicedema,thebloodbrainbarrierBBBisimpaired,leadingtoaninuxofintravascular uidsandproteinsintotheCNS.TheedemaspreadsquicklyintotheCNS,movingextracellularly alongwhitematterbertracts. [66] The'diffuse'descriptionofswellinginvasogenicedemarelates toitsprogressionthroughthebrain,asopposedtoaheterogeneousexpressioninvolvingmultiple independentregions.Theintracellularspacebeginstollupwiththeadditionaluids,causing swellingwithintheregions.Vasogenicedemacanalsospreadtogreymattertissuesaswell.CytotoxicedemadoesnotinvolvepermeabilityoftheBBB,butinsteadresultsfromdisregulationofion pumpactivity,leadingtovolumesofwaterbeingtakeninsidethecells,whichcausesthecerebral swelling.Adysregulationoftheiongradientintheintracellularspacecreatesamicroenvironment thatisnotsuitableforneuralimpulsering,andcellsoftenundergoanecroticcelldeath.Ionpump 14

PAGE 20

dysfunctionncanoccurthroughavarietyofmechanisms.Inregionssufferingischemia,impoverishedsupplyofglucoseandoxygentobraincellssuspendsfunctionofmembranesodiumand calciumionpumps.Asubsequentincreaseinthesodiumandcalciumcontentoftheintracellular spacecausesnearbycellstoabsorbexcessquantitiesofwater. [114] Alternatively,metabolicchanges cancauseglialcellpotassiumandsodiumpumpstosuspendtheiractivity,causingincreasedretentionofwaterandsodiumionswithinthecell.Astrocytesaremorepronetoswellingincytotoxic edemaduetotheirroleinionre-uptake. [66] Thisstateofretentionpersistsevenwhenanaction potentialwouldnormallyreleaseintracellularsodiumionconcentrations.Thisretentionleadsto similarswellingfromincreasedwatercontent.Glialcellionpumpdysregulationhasalsobeen knowntobetriggeredfromexposuretoavarietyoftoxins. [44] Brainherniationoccurswhensevereintracranialpressureshiftsbrainpositionandputsstrainon aportionofbraintissueasitispressedagainsttheskullorotherextra-axialstructures.Displacement canbecausedbyhematoma,tumor,orintracranialhemorrhage.Brainherniationcanalsooccuras aresultofTBI,eitherbyproducinghematomaorhemorrhagingthatcausesdisplacement,orsimply fromtheforceofthetrauma. [104] Becausetheregionbeingcompressedmaysuffersignicantdeformation,ischemiaofthetissuesisverycommon,andoftenresultsininfarction.Herniationcanoccur inalmosteverybrainregion,butthethreemostcommontypesofherniationarethecingulateAKA 'subfalcine',uncal,andtonsillarherniation. [53] Cingulateherniationiswhenthecingulategyrusis compressedagainstthefalxcerebri,theduramaterseparationoftheinterhemisphericssure. Figure4:SagittalviewofthefalxcerebriasdepictedbyGray'sAnatomy [115] 15

PAGE 21

Uncalherniationcausestheinnermostportionofthetemporallobetobecompressedtowards thetentoriumcerebelli,puttingpressureonthebrainstem.Finally,tonsillarherniationisacompressionofthelowercerebellartissuesandtractsthroughtheforamenmagnum,theopeningforthe medullaoblongataandbrainstemrunningoutoftheskull.Thisputssignicantradialpressureon thebrainstem,andcandisruptnormalrespirationandautonomicnervoussignaling. [75] Infarctionissimplythenecrosisoftissuesthathavebeendeprivedofoxygen,usuallybyischemia.Infarctionitselfcanserveasastructuralimpedimenttobloodowandcirculationof oxygentolivingtissues,leadingtomoresevereischemia.Infectionofbraintissuescancasesof localizedinfectionsspreadingpasttheBBB.Infectedbraintissuesdevelopintoalesionofnecrotic infectedtissue,increasingintracranialpressure,anddisruptingfunctionoftheregionwhereitdevelops.Althoughantibioticsorsurgerytoremovetheabscesscanreducethelong-termdamage potentialofinfection,thereisnomethodforrecoveryofinfectedtissues. Chronic Anysecondaryinjurywithanonsetthatisdelayedrelativetotheinitialtrauma,orbecomesprogressivelyworseifleftuntreatedisconsideredchronic.ThisincludesHydrocephalus,encephalomalacia,cerebrospinaluidleak,leptomeningealcyst,walleriandegeneration,andchronictraumatic encephalopathy. HydrocephalusisthebuildupofexcessCSFintheventricles.Normally,CSFcirculateswithin theopenbrainspacesanddrainsfromtheventriclesintothecirculatorysystem,unlesstheforamina ofthefourthventricleorthecerebralaqueductareblocked. [112] BuildupofCSFintheventricles becomesaproblemwhenitalreadyllstheentirespaceavailable,andmoreisbeingproduced. Theinternalpressurebuildupinhydrocephalusissimilartothatofintraventricularhemorrhaging. Symptoms,damageproles,andtreatmentsareverysimilarforboth. [108] Hydrocephalusisusually acongenitalmalformationordevelopsfromtraumaandsubduralhemorrhaging.Thistypeofcomplicationismostcommonininfantsandveryyoungchildren,andislesslikelytooccurincasesof adolescentoradultmTBI. [112] Encephalomalaciaisasofteningofbraintissues,andiscommonlycausedbyprolongedbuildup ofintracranialblood.Theconditionissomewhatdegenerative,andtheaffectedtissuesstopfunc16

PAGE 22

tioningentirelywhentheybecomesoft.Braindamagecanbesevereandlocalizedbothtogrey orwhitematter,andnoknowntreatmentsareabletorepairthesoftenedtissues. [9,69] Thistypeof complicationcanbeverydangerous,butluckilyitonlyoccursinaminorityofcasesofmoderatetosevereTBI.mTBIisunlikelytodevelopencephalomalaciaunlessitactsasatriggerfora pre-existingcondition. Leptomeningealcystisaskullfracturethatalsohassomeduraltearing.Thisleadstoherniationofthepiamaterandarachnoidlayersthroughtheduraltear.Thebonelayerexperiencing fracturewilloftenexperienceerosionfromthevascularactivityofthemeninges.Theonsetofleptomeningealcystsarealmostalwaysininfantsandveryyoungchildren,whentheskullplatesare stillrelativelyweak. [113] WallerianDegenerationisthedegenerationofanerveberandresultingaxonaldeaththathas beenseparatedfromitsnutritivecenterbyinjuryordisease,characterizedbybreakdownofthe myelinandresultinginatrophyanddestructionoftheaxon. [18] Thispathologicalprocessbegins witharapidaxonaldisintegrationandbreakdownofmyelinsheath,thenactivationofmicroglia, withsubsequentclearanceoftissuedebris.Effectivelyaxonaldegeneration,microglialactivation, andnallyastrocytosisarealllabeledaspartsoftheprocess"Wallerian". ChronictraumaticencephalopathyCTEisaneurodegenerativeconditiontriggeredinpartfrom repetitivemildtraumaticbraininjuriesorconcussion.Thewalleriandegenerationobservedinnormalmicroglial'cleanup'followingmTBIdoesnotshutoffwhenappropriate,andamaladaptive feedbackloopisentered. [2] InCTE,Brainregionscontinuetoexperiencedegenerationevenin theabsenceoffurthertrauma,andcognitivesequalaeof'punchdrunkenness'develop. [28] CTEis thoughttoresultinwidespreadcognitivedysfunctionandpatientsmayexhibitimpulsiveness,memorydecits,personalitychanges,apathy,depression,anddementia.Beyondrepetitivebraintrauma, otherriskfactorsfordevelopingCTEremainunknown.Accumulationofhyperphosphorylatedtau andTDP-43havebeenobservedaspotentialmolecularmarkersforthepresenceofCTE. [2] Unfortunately,CTEcanonlybediagnosedwithcertaintypost-mortemsinceinvivoimaginghasfailedto producepredictivebiomarkers. 17

PAGE 23

3SummaryofVariousNeuroimagingModalitiesinmTBIResearch SomemedicalimagingtechniqueslikeX-raystillproducetwodimensionalimages,butmost modernimagingnowinvolvestheusethreedimensionalspatialrepresentationsfortheirdata.Performingproperinvivoneuroimagingcertainlyrequirestheresolutionofallthreespatialdimensions toobserveanyregionofthebrain.Thethreedimensionalimagesusedarecomposedof'voxels'. Similartothewayinwhichapixelisthesmallestelementofatwodimensionalpicture'sdata, avoxeldescribesthesmallestelementofdatainathreedimensionaldataset.Avoxelisathree dimensionalpixelwithwidth,height,anddepth.Asinglevoxelcoversacubicspace,andcontains somevaluethatismeanttodescribeinsomewaytheunderlyingcubicspace.Avoxelcouldcontainvaluesregardingglucoselevels,ortissuedensity,orwaterdiffusionproperties,aslongasthe spatialmapofvoxelsrepresentsthesequalitiesnumerically. [73] Inthischapterwewillexaminea varietyofneuroimagingmodalitiesthatusevoxels,andhowtheycandetectanddiagnosemTBI usingmethodsofstatisticalinference.Thischapterwillcoverabriefliteraturereviewofvarious imagingstudiesofmTBIusingimagingmodalitiesotherthantheonesusedinthisthesis.Review ofsMRIandDTIliteraturendingswillbeconductedinthechaptersthatfollow. 3.1ImagingModalitiesforDetectionofmTBIPathology InordertodeterminetheutilityandfunctionalityofeachimagingmodalityfordetectingmTBI, itisparamounttorstobservetheresultsofthecurrentliterature.Inthissection,asummarydescriptionofvariousimagingmodalitiesandhowtheyhaveperformedonmTBIdetectionisprovided. AlthoughthemajorityofliteraturesupportstheuseofsMRIorDTImodalitiesasmosteffective indetectionofDAI,variousalternatemodalitiesmayrevealotheruniqueaspectsofmTBIpathology.ThesealternatemodalitiesmayhelpdevelopabetterunderstandingofoverallmTBIpathology whenperformedinconjunctionwithsMRIorDTI.Anumberofmeta-analysesandtheirndings aredescribed,andthemyriadreferencesandrecommendationsfromthesemeta-analysesareused asaprimarysource,withadditionalstudiesprovidedtosupportorcontesttheirassessments. Forexample,inTuongH.LeandAlisaD.Gean,theauthorsdetailabroadoverview ofneuroimagingtechniquesforanalysisofTBI.TheirreviewoverallrecommendstheuseofsMRI onlyforpatientswithsub-acuteorchronictraumaticbraininjury,butrecommendsdiffusion-weighted 18

PAGE 24

imaging,magneticresonancespectroscopy,andmagnetizationtransferimagingforuseindetecting mildtraumaticbraininjury,whichotherimagingtechniquesarelessabletodiagnose.[1] Similarly,inBovenetal.,theyreviewthecurrentstateofvariousneuroimagingmodality studiesandcomparethemwithDTIstudiesidentifyingbiomarkersoftraumaticbraininjuryand post-traumaticstressdisorder.Thereviewincludessummariesofsevenotherarticleswhichused diffusiontensorimagingtoinvestigatemildtraumaticbraininjury.Mostofthesestudiesused datasetsof10-20subjects,andselectedsubjectsprimarilybasedonthetimesincetheincidenceof theirinjury. [92] SkullFilms Tobeginananalysis,askullX-raymaybetakenasarststep.Thisimagingmodalityprovidesan imageoftheouterskull,andisoftenusedtodetermineiffractures,duralpenetration,orsignicant swellingoccurs.TheseinitialmarkerscanindicatethepresenceofsevereTBI,andhelpquickly localizeregionsatriskforhematomaandhemorrhaging.However,becausetheydonotprovide insightbeyondsupercialstructures,theyremainpoorpredictorsofintracranialpathology.Skulls lmsandotherX-raytechniquesarelimitedintheirabilitytoperformrealinvestigativeinvivo imagingofthebrainandintracranialtissues.Itiswidelyagreedthatskulllmsshouldnolonger beusedinclinicalevaluationofofclosedheadTBI,especiallysinceCTscansdoamuchbetter jobofalertingclinicianstodangeroushematoma. [1,99,10] CTbegantoreplacex-rayasthestandard practiceforheadinjuryevaluationduringthe1980s. [79] ComputedTomography ComputedtomographyscansCTscansareutilizedasastandarddiagnosticininitialinvestigationforavarietyofinjurytypes.CasesinwhichCTscansarecommonlyperformedinclude: moderateandsevereTBI,patientswithagegreaterthan60years,persistentneurologicaldecit, chronicheadacheorvomiting,amnesia,lossofconsciousnesslongerthan5minutes,depressed skullfracture,penetratinginjury,orbleedingdiathesisormeasuringtheeffectsofanticoagulation therapy. [89] Thereasonthismodalityisaclinicalstandardprocedureisbecauseitisfast,widelyavailable, andhighlyaccurateinthedetectionofskullfracturesandintracranialhemorrhage.CTscansare 19

PAGE 25

almostalwayscompletedwithin5minutes.Actualscantimeislessthan30secondsonaverage. [87] Therefore,CTisalsolesssensitivetopatientmovementthansMRI.CTcanbesuperiortosMRIin revealingskullfracturesandradio-opaqueforeignbodies,whilesMRIremainscomparabletoCT indetectionofacuteepiduralhematomaandsubduralhematoma. [79] Incontrasttotheseapparent superioritiesincertainareas,CThassomeissuesthatmakeitundesirableasastandardimaging modalityinclinicalsettings. [87] Forexample,beam-hardeningeffect,whichisthedisplacement oftheCTsignalnearmetalobjects,bone,calcications,andhighconcentrationsofcontrast,can degradetheimagequalityandpreventaccurateassessments.CTcanmisssmallamountsofblood thatoccupywidthslessthanaslicebecauseofvolumeaveraging.CTndingsmayalsolagbehind realintracranialdamage,sothatscansperformedwithin3-8hoursoftraumamayunderestimate injury. [90] EvenwithsignicantdevelopmentsinCTimagingoverthepreviousthreedecades,themajority ofconrmedmTBIcasesstillshownovisibleabnormalitiesonCTscans. [87,1,92,90] Despitethis, CTisstillwidelypracticedinresponsetoheadinjury,andanyfollow-upimagingwhentheneurologicalndingsareunexplainedbytheCTndingsisprimarilydonebysMRI.sMRIremainsthe preferredclinicalimagingmodalityforsub-acuteandchronicTBI.Unfortunately,lackofndings ininitialCTscanswillsometimespreventfurtherinvestigationusingamoreappropriatemodality. Gradient-Recalled-EchoGRET 2 *-WeightedMRI T 2 *-weightedMRI,commonlyreferredtoasGRE,ishighlysensitivetothepresenceofferritin andhemosiderinmaterials,whichareleftoverproductsproducedinbloodplasma.Bothcompounds alternearbytissue'smagneticsusceptibility,whichishowthepresenceofthesematerialsisdetectedbythesensors.T 2 *imaginghasbeenlimitedinthedetectionofcorticalcontusionsinthe inferiorfrontalandtemporallobesbecauseoftheinhomogeneityartifactsinducedbytheparanasal sinusesandmastoidaircells. [1] However,otherstudieshaveshownthatT 2 *-weightedimagesperformsignicantlybetteratdetectinglesionsinTBIthannormalT 2 scans,andexhibitedveryhigh correlationbetweeninjuryseverityandbothlossofconsciousness,andscoreontheGCS. [74] These ndingssupporttheuseofT 2 *imagingfordetectionofheadinjuryinvolvinghemorrhaging,but unfortunatelymanycasesofmTBImaylackevenminorbleeding.ThereforeT 2 *imagingisnot 20

PAGE 26

anidealmodalityfordetectingallcasesofmTBI,onlythoseinjuriesexpressinghemorrhageor bleeding. Diffusion-WeightedImagingDWI Diffusion-weightedimagingdetectsrandommotionofmoleculesinbraintissues.Itcanbe usedtodetectregionalchangesinneuronmembranepermeabilitytoNa+andK+ions,aswellas intracellularwaterpermeability.InDWI,theintensityofeachvoxelreectsthebestestimateofthe rateofdiffusionalongaspecicvectorwithinthatspace.ThisapparentdiffusioncoefcientADC producedfromDWIimagingmeasuresthemagnitudeofdiffusionthroughthevoxel.Molecular diffusionofions,proteins,andwaterintissuesisnotfreeinalldirectionsisotropic,butreects interactionswithmanyobstacles,suchasmacromolecules,bers,membranes,etc.Watermolecule diffusionpatternscanthereforerevealmicroscopicdetailsabouttissuearchitecture,ineithernormalordiseasedstates.Diffusioncharacteristicswithingreymatteraremostlyisotropic,whilewhite mattertissuesdisplaymuchgreateranisotropyintheirdiffusionproperties.ThereforeDWIprimarilyrevealschangesinwhitematter.Becausethemobilityofwaterdiffusionisdrivenbythermal agitationandishighlydependentonaspectsofthecellularenvironment,thehypothesisbehind DWIisthatndingsmayindicateminuteearlypathologicalchanges.Forinstance,DWIismore sensitivetoearlychangesfollowingstrokethantraditionalMRIT 1 orT 2 -weightedimages. [92,1,18] Additionally,measuresoflateralADCwithininterhemisphericcommissuressuchasthesplenium ofthecorpuscallosumhavebeenshowntodecreaseincasesofchronicmTBI [17,56] MagnetizationTransferImagingMTI MagnetizationTransferImagingexaminesthelongitudinalT 1 relaxationrelationbetweenbound protonsandbulkfreewaterprotons.Whenaradiofrequencypulsethatisoffresonancewiththe magneticeldisapplied,itselectivelyexcitesthoseprotonsthatareboundinmacromolecules. Theseprotonssubsequentlyexchangetheirexcesslongitudinalmagnetizationwithnearbyfreewaterprotons.Themagnetizationtransferratioprovidesaquantitativemeasureofthestructuralintegrityoftissue.AreductionofthemagnetizationtransferratiocorrelateswithworseclinicaloutcomeinapatientwithahistoryofTBI. [1,108] InvestigatorshaveusedMTItodetectwhitematterab21

PAGE 27

normalitiesassociatedincasesofmultiplesclerosis,progressivemulti-focalleukoencephalopathy, andWalleriandegeneration. [1] MTIchangeshavebeenfoundtobemoresensitivethanT 2 -weighted MRIindetectinghistologicalaxonaldamageinanimalmodels. [95] SinglePhotonEmissionTomographySPECT SinglePhotonEmissionTomographyprimarilymeasurescerebralbloodowCBF.Overall,fMRIprovidesagreaterspecicityandresolutionthanSPECTindeterminingbloodusedeoxygenationinthebrain.However,SPECTdetectstheactualmovementcharacteristicsofblood throughoutthebrain,regardlessofitsuse,whichcanrevealabnormalitiesorobstructionsthatinhibit bloodsupply.SPECTmaybeabletofunctionasalongtermprognosticindicatormoreeffectively thanCTorMRI.WorseprognosishasbeenassociatedwithmultipleCBFabnormalities,larger CBFdefects,andSPECTdefectsthatinvolvethebasalganglia,temporalandparietallobes,and brainstem. [86] SPECTislesssensitivetothedetectionofsmalllesionsandDAIthatarepotentially visibleinsMRI. [38,86] Therefore,SPECThasfunctionalpotentialfordetectingabnormalbloodow whenusedtocomplementastructuralimagingmodalitysuchassMRIorDTI,butSPECTisnot viableasastandalonemethodforevaluationofTBI,andespeciallymTBI. [1] PositronEmissionTomographyPET Positronemissiontomographyisanimagingmodalitythatdetectsgammaraysemittedindirectlyviaatracerthatisintroducedintoasubjectscirculation.Abiologicallyactivemoleculeis taggedwithaprotonemittingradionuclide,whichgivesoffgammarayswhenitisprocessedor brokenapart,toobservewherethatmoleculeisbeingutilizedinthebody.PETneuroimagingcommonlymeasuresglucoseasthebiologicallyactivemolecule,whichrevealsbrainmetabolismof glucose.FluorodeoxyglucoseFDGisusedasthetracer,sinceitisaglucoseandemitsgamma rayswhenprocessed.Theconcentrationsoftracerimagedwillindicatetissuemetabolicactivityby virtueoftheregionalglucoseuptake. [40,85] Glucoseinthebrainisoftenprocessedabnormallyin damagedregionspostTBI. [1] Inanimalstudies,acutelyinjuredbraincellsshowincreasedglucosemetabolismfollowingsevereTBIduetointracellularionicperturbation.Followinginitialhyperglycolysis,injuredbrain 22

PAGE 28

cellsshowaprolongedperiodofregionalhypometabolismlastinguptomonths. [44] However,PET studieshavehadlimitedsuccessdemonstratingconsistentresultsinlocalizinganddifferentiating mTBIinhumans.BecauseoftheheterogeneousnatureofTBI,studieshavefoundbothhypermetabolismandhypometabolisminthesameregionsacrossdifferentpatients. [85,92] Therefore,itis importanttoconsiderthatmetabolicabnormalitiesinmTBIarelikelynotlimitedtoglucoseuptake inproximitytoapparentlesions.InmTBI,metabolicdysregulationofotherproteinsandmaterials mayoccurmoresignicantlythanglucose,butlittleresearchexistsforothermetabolites.While PETmayshowsomeindicationsofgeneralizedmetabolicchangeincasesofmTBI,theyarenot entirelyco-localizedwithdiffuseaxonalinjuryorindicativeoftheregionsaffectedbyasubject's injury. [92,40] 4T 1 -WeightedMagneticResonanceImagingsMRI T 1 -weightedMRIisatypeofstructuralimagingwhichdisplayssignicantcontrastbetween thegrayandwhitemattertissuesofthebrain.T 1 -weightedimagingisthecanonicalstandardfor neuroimagingmodalitiescurrentlyinuse,andiscertainlyoneoftheoldest.Itismorecommonly referredtoasstructuralMRIsMRI. 4.1DevelopmentandMechanics Allmagneticresonanceimagingreliesontheexcitabilityofprotonswithinthenucleiofatoms,and theratesatwhichittakestheprotonstodispersetheirexcessmagnetizationtonearbyprotonsand returntotheirnaturalstate. [118] Allatomicnucleihavedistinctcharacteristicsofnuclearspin,andmagneticmoment.The nuclearspinquantumnumberofanelement,representedbythesymbol I ,describesthetotalangular momentumofitsnucleus. I isusedtodeterminehowmanyspinstatesanucleuscanhave,usingthe followingfunction:2 I + 1=allowedspinstates.Thenuclearspinstatedescribesthedirectionsof thenuclei'srotationaboutanaxis,whichisthemagneticmoment.Nucleithathaveevennumbersof protonsandneutronswillequalize,andthereforehavenospinstates,with I = 0.Nucleicontaining anoddsumofprotonsandneutronshavespinstatesof 1 2 integerincrements,andnucleicontaining oddnumbersofbothprotonsandneutronswillhavespinstateswithintegerincrements,2,etc.... InsMRI,weareinterestedinobservingthenucleusofhydrogenatoms,whichhasanuclear 23

PAGE 29

quantumnumberof I = 1 2 .Therefore,hydrogenatomshaveatotalof2possiblespinstatesofequal angularmomentum, + 1 2 and )]TJ/F44 7.9701 Tf 9.672 4.298 Td [(1 2 ,whichrepresentclockwiseandcounterclockwiserotationaround themagneticmomentrespectively.Thereexistsa'lattice'ofmagneticenergyintheenvironment surroundinganucleus.AmbientnoisyenvironmentalenergiessuchasheatandcertainEMwavelengthscauseconstantrotationandmovementofthenucleusasitattemptstomaintainamagnetic momentofleastinterferencetoitsangularmomentum. [117] Anyshiftinnucleusmagneticmoment orpositionrelativetothesurroundinglatticecausesthenucleustoproduceamagnetic'latticeeld' ofitsown.Thelatticeeldofanucleusiscapableofdiffusingitsexcessmagnetizationintothe latticeeldsofothernearbynucleiwithlowerexcitation,viamagneticresonance.Thelatticeeld ofanexcitednucleuswillnaturallyattempttoreturntoastateofminimalenergybydistributing itsexcessmagnetizationtothelatticeeldsoflowerenergynuclei.Eventually,nucleitransferof excessenergieswillachieveabestcase'restingstate'latticeofthermodynamicequilibriumrelative toneighboringlatticeelds. [119] Whenthisisaccomplished,themagneticmomentofeachnucleus israndomlyaligned. Whenhydrogennucleiareexposedtoastaticmagneticeld, B 0 ,thatisstrongerthantheir surroundinglatticeeld,theybegintoactlikedipolemagnetsandaligntheirmagneticmoment alongthesameaxisasthestaticmagneticeld.Becausetherearetwospinstatesforhydrogenthis alignmentcaneitherbeup-spin,alongtheexternaleldorientation,ordown-spin,againsttheeld orientation.Thenucleiwillnotalignperfectlywiththestaticeld B 0 ,butinsteadwill'precess'or wobblearoundtheaxisastheyspin.Thenucleiwillallprecessatthesamefrequency,butaredoing soatoutofsyncwithoneanother.WhenbombardedbyradiofrequencyRFenergy,nucleiwill absorbsomeofitifthefrequencyusedisresonant,anditwillbeabletoaltertheirmagneticmoment andspin. [121] ThefrequencyofRFthatisresonantwithanucleusiscalledtheLarmorfrequency, givenbythefunction: w = B 0 g Where g isthegyromagneticconstant,and B 0 isthemagnitudeofthestaticmagneticeld. TheLarmorfrequencyisalsothefrequencyatwhichtheprecessingofnucleimagneticmomentis occurringaroundthe B 0 axis.Inhydrogennuclei,fora1.5Teslaeldstrength,theresonantRFis 24

PAGE 30

63.9MHz. [117] Inordertobringprecessingnucleiintosynconthesamex/yplane,asecondeld emittingtheresonantRF, B 1 ,isappliedperpendicularto B 0 ,androtatedaboutitsaxisastheRF energyisadded.Oncethenucleiareprecessinginphasetogether,a B 1 pulseisappliedalongthe x/yplane,andperpendiculartothe B 0 axis.Thisallowsachangeintheangleofthethenuclear magneticmomentduetoexcessmagnetizationonthex/yplaneandthezaxis.Inorderfornuclei toreturntotheiroriginalmagneticmomentbyenergytransfer,theexcitednucleimustbeexposed toalatticeeldoscillatingatafrequencyatorclosetotheLarmorfrequency.Thusthespeedat whichthisrelaxationoccursinanucleusisdependentonthenumberofsurroundingnucleithatare precessinginsyncandattheLarmorfrequency.Thisrelationshipisassumedtohaveanexponential behavior. 4.2MathematicsandMeasures T 1 isthelongitudinalrelaxationtime,alsoknownasthespin-latticerelaxationtime.Itrepresents theamountoftimeittakesforanucleimagnetizedata90angletoreturnto63%ofitsoriginal magneticmoment.T 1 isgivenbythefollowinggeneralequationasderivedfromBlochequations forthenuclearmagnetizationofallthreespatialvectors M = f M X ; M Y ; M Z g : M z t )]TJ/F94 10.9091 Tf 9.991 0 Td [(M 0 =[ M z 0 )]TJ/F94 10.9091 Tf 9.992 0 Td [(M 0 ] e )]TJ/F94 5.9776 Tf 4.495 0 Td [(t T 1 M z t = M 0 )]TJ/F15 10.9091 Tf 9.991 0 Td [([ M 0 )]TJ/F94 10.9091 Tf 9.992 0 Td [(M z 0 ] e )]TJ/F94 5.9776 Tf 4.495 0 Td [(t T 1 Inthecasethat M z 0 = 0,byensuringthatallnucleiaresyncedandprecessingonthesame x/yplane,theequationcanbereducedto: M z t = M 0 1 )]TJ/F94 10.9091 Tf 9.991 0 Td [(e )]TJ/F94 5.9776 Tf 4.495 0 Td [(t T 1 Where: M Z =Theorientationofnuclearspinmagnetizationonthezaxis. M 0 = magneticmomentorientationatthermalequilibrium g = )]TJ/F94 7.9701 Tf 6.193 0 Td [(charge 2 mass = Thegyromagneticconstant B 0 = Staticmagneticeld=.5-Tesla B 1 = ResonantRadiofrequency=63.9MHzat1.5-Tesla w = B 0 g = TheLarmorprecessingfrequency 25

PAGE 31

5DiffusionTensorImagingDTI Diffusiontensorimagingisamodalitythatdescribesthe3dimensionaldiffusionproperties withineachvoxelasrepresentedbyamatrixofeigenvaluescalledatensor.Itshowsgreatpotential fordetectingdisturbancesinwhitematterberbundletractintegrity,andmaybemoreeffective thansMRIfordetectingdiffuseaxonalinjurythattakesplaceinmTBI. 5.1DevelopmentandMechanics Priortothedevelopmentofdiffusiontensorimaging,themeasureofdiffusionwithintissueswas onlyobservablebydiffusionweightedimagingDWI.Itwasonlyrelativelyrecently,starting around1992whentherstimageswereproduced,thatdiffusiontensorimagingbecameanavailable methodofresearchandinvestigationinthebrain. [80] DWIalonecanonlydetectthediffusionpropertiesofavoxelalongasingledirectionalinuence gradientatatime,andsodiffusiontensorimagingwasdevelopedtorepresentmultipledirections ofdiffusionsimultaneously.Ifdiffusiondataarecollectedinatleastthreeorthogonalplanes,thena threedimensionaltensorcanbecalculated.Additionaldiffusiondirectionscanbeobtainedtoproducetensorswithbetterrepresentationofvaryingdiagonalandnon-planardiffusionproperties.The propertiesandstructureofindividualaxonscannotbedescribedusingDTI,butdiffusionmeasures candescribepropertiesoftheunderlyingtissuesthatareinformativeastothegeneralcharacteristics ofneuronberbundles. [17] 26

PAGE 32

5.2MathematicsandMeasures InordertoobtainDTIdata,anumberofinitialscansandcomputationsmustbemade.Acquisition ofanMRIvolumeforspatialoverlayisfollowedbysixDWIvolumessometimesmore,eachwith adifferentmagneticgradientappliedfromnon-collineardirections.EachDWIimageisessentially measuringthediffusioncharacteristicsinaspecicdirectionforeveryvoxel.Eachgradientcauses adifferentdirectionofdiffusiontobemeasuredfortheimage.Thecomputationofthemagnitude ofdiffusioninasingledirectionforavoxelisgivenbythefollowingequation: I q = J e )]TJ/F94 8.9664 Tf 6.967 0 Td [(b q D Where: I q = Theimagevoxelintensityforeachgradientq J = Theidealimageintensitywithoutanappliedgradient,usuallysetas J = I 0 D = timedifferencelagbetweenstartsofthegradient d = durationofthegradient g = )]TJ/F94 7.9701 Tf 6.193 0 Td [(charge 2 mass = Thegyromagneticconstant G = thestrengthofthegradientpulse b q = theb-matrix= g 2 G i G j d 2 D )]TJ/F100 10.9091 Tf 9.991 0 Td [(d = 3 fortheqthencodinggradient Thediffusiontensorcanthencalculatedlikeso: ln I 0 = I q = b q D D = ln I 0 = I q b )]TJ/F44 7.9701 Tf 6.193 0 Td [(1 D=Thediffusiontensor,asymmetricmatrixrepresentingdiffusioninatleast6principaldirections: Thediffusiontensorcanberepresentedasathreedimensionalellipsoidwhoseshapedescribes therelativemagnitudesofdirectionaldiffusion. Figure5:Ellipsoidrepresentationsofanisotropicandisotropicdiffusiontensors. 27

PAGE 33

FAfractionalanisotropyisascalarmeasurefrom0to1ofthedirectionalityofdiffusionat avoxel.Itdescribestheseverityoftheelongatedshapeinatensorellipsoid.InCSFandgray matter,waterdiffusionisclosetoisotropic,andsotheFAiscloseto0.Inwhitematter,suchasin thecorpuscallosum,thewaterisrelativelyfreealongtheaxons,butrestrictedperpendiculartothe axonsandthereforemoreanisotropic,withFAbeingcloserto1.Thusinwhitematter,reducedFAis generallythoughttoreectlossofwhitematterintegritythatmayreectdamagetomyelinoraxon membranedamage,orperhapsreducedaxonalpackingdensity,and/orreducedaxonalcoherence. SuchdamageisconsistentwiththemicrostructuralchangesthatoccurinDAI.FAismeasuredas thedegreeofdifferencebetweentheeigenvaluesofthediffusionmatrix,computedas: FA = p l 1 )]TJ/F100 10.9091 Tf 9.991 0 Td [(l 2 2 + l 1 )]TJ/F100 10.9091 Tf 9.991 0 Td [(l 3 2 + l 2 )]TJ/F100 10.9091 Tf 9.992 0 Td [(l 3 2 p 2 q l 2 1 + l 2 2 + l 3 2 MDMeanDiffusivityisascalarmeasurefrom0to1oftheapparentdiffusioncharacteristic, averagedfromtheprimarydirectionsofdiffusionwithinavoxel.Itdescribestheoveralleaseof diffusion,orthemean'size'ofthediffusionwithinavoxel.ThemaximumMDof1isrepresentative ofcompletelyunimpededisotropicdiffusion,whilelowervaluesimplyadensersubstratewithless overalldiffusion,evenifitremainsisotropicinshape.MDisoftenincreasedinareasthathave sufferedcelllossandasubsequentincreaseinextra-cellularspace.MDrepresentstheaverageof theDWIapparentdiffusioncoefcientsinall3spatialdirections,andiscomputedbythefollowing equation: MD = l 2 1 + l 2 2 + l 3 2 3 RadialDiffusivityRDandAxialdiffusivityADarealsofairlycommonmeasuresofdiffusion propertiesobservedinDTIresearchstudies. [15,1] Theyrepresentthemagnitudeofdiffusionwithin avoxelonlyonahorizontalplaneorverticalplane,respectively.However,thesemeasureswillnot beusedinthisstudy. 28

PAGE 34

Tractographyisa3Dmodelingtechniqueusedtovisuallyrepresentneuraltractsusingdata collectedbyDTI.Whentheprimaryvectorofananisotropicvoxelissufcientlyalignedinterms oforientationwithaneighbor'sanisotropicorientation,atractisdrawnconnectingthetwovoxels. Ifatractisproduceditindicatesthepresenceofacontinuousbundleofwhitematterbers.The varyingdirectionalityofthesetractscanberepresentedaswellforeachlayeroftheimage,using colorcodestodescribeplanardirectionality. Figure6:AnexampleofDTItractographyoverlayedontoaT 1 structuralimage [69] 5.3DTIdetectionofmTBI Bovenetal.havedemonstratedthatFAissignicantlyeffectiveinidentifyingdiffuseaxonal injuries,whichtechniquessuchasstructuralMRIandCTscanshavedifcultyidentifying. [92] Previouscross-sectionalstudiesexaminingtheevolutionofanisotropicdiffusioninsubjects acrossmulti-yearperiodshavenotreportedsignicantdifferencesbetweensemi-acutemTBIand patientswithchronicmTBI. [57] Thisislikelybecausethegreatestdiffusionchangesoccurduring acutemTBI,andthattheDTIchangesobservableinchronicmTBImaybesignicantlydifferent fromhealthycontrols,butnotfromsubacutemTBIpatients. Shentonetal.provideoneofthemostcomprehensivecompilationsofsMRIandDTI 29

PAGE 35

studiesonmTBI.Eachofthe43DTIstudiesinvestigatingmTBIreportedsomeDTIabnormalities. However,anatomicallocationofeffectswererarelysimilar.Thelackofsimilarregionsimplicated intheirresultsisnotsurprisingconsideringtheheterogeneityoftraumaticbraininjuries,aswellas thelargevariabilityinthepresentedstudiesbetweentimeofinjuryandDTIscan.Someregions, however,arereportedmoreoftenthantheothers,apatternwhichmightsuggesttheirincreased vulnerabilitytoaxonalinjury. [15] AmajorconsiderationobservedinShentonetal.'sreviewisthe differencesinacuteandchronicmTBIdiffusionpropertiesreportedbystudies.Toconsiderthe abnormalFAincreasetheyfoundinsomeacutemTBIstudies,itisimportantthatfutureinvestigationsdetermineifFAincreaseshortlyfollowingheadinjuryiscorrelatedwithworselong-term prognosis.GiventhatsomeinvestigatorsinthereviewdonotreportincreasedFAat24hours post-injury,whileothersreportincreasedFAat72hours,itisunclearwhatconstitutesanaverage or'normal'progressionofFAchangefollowingTBI.Followupresearchshouldconsiderrepeated imagingofpatientsduringtheacuteandsubacutephases,soastoobservetheprogressionofFAand MD.SubjectswithincreasedFAanddecreasedMDintheacutephaseofinjuryismorecommon inthestudiesobserved,anddecreasedFAandincreasedMDwasmorecommoninchronicmTBI patients. [56,51] Thiscouldimplydifferentcharacteristicsoftissuedamageresponseandswelling earlyon,comparedtothelonger-termstructuralchangesobservedin>3monthpost-injuryimaging.Morelongitudinalstudiesperformingscansatmultipleintervalsoverthecourseof3months postinjuryareneededtoelucidatetheeffectsofacuteandchronicmTBIprogressionprolesofFA andMD. Rutgers,etal.comparedFA,ADC,andtractographywithinROIsbetweengroupsof 11healthycontrolsand21mTBIpatients.Theirresultsshowedthatonaverage,patientsshowed reducedFAin9regions,predominantlyincerebrallobarwhitematter,cingulate,andCCcompared tocontrolsfor191regions.Fortheirtractographyobservations,discontinuityinbertracking wasseeninonly19.3%ofbers,indicatingameasurableeffect,butlikelylimitedtolargerber bundles. [33] 30

PAGE 36

Figure7:DiscontinuityintractintegritymayindicatethepresenceofDAI [34] Inafollowupstudydescribedinthesamepaper,andusingthesamesubjects,Rutgers,etal. examinedthedifferencesofacuteandchronicmTBIintermsofFA,MD,andADCmeasures. PatientswithmTBIscannedlessthan3monthspost-injuryshowedreducedFAandincreasedADC inthegenuofthecorpuscallosum,whereaspatientsscannedafter3monthspost-injuryshowedno suchdifferences.TheyalsoshowedthatmoreseveretraumawasassociatedwithFAreductioninthe genuandspleniumofthecorpuscallosum,alongwithincreasedADCandfewernumbersofbers. Theirndingssuggestthattheremaybeareversalofdamagetothecorpuscallosuminpatients withmTBIwhorecoverafter3months,andthatmoreseveredamagetothecorpuscallosummay beassociatedwithaworseoutcome. [33] Huismanetal.conductedastudyinvestigatingpotentialbiomarkersofwhitematter forimprovingpredictionofmTBIusingDTI.Twentypatientswereevaluatedandcomparedwith fteenhealthycontrolsubjects.ADCandFAvaluesweremeasuredatmultiplepreselectedregions ofinterestROIsandcorrelatedwithclinicalmeasuresofTBIseverity.TheyincludedTBIpatients withGlasgowComaScalescoresbetween4and15,thereforemoderatetojustbelowseverewere notseparatedfrommildinevaluation.WhilethissamplemaynotberepresentativeofanmTBI population,thegoalofthestudywasinparttodeterminethecorrelationoftheglasgowcomaand dischargeRankinscoreswithseverityofDAI.Theirresultsconrmthatchangesinwaterdiffusion anisotropyoccurinTBI.FAwassignicantlydecreasedintheposteriorlimboftheinternalcapsule meandecreaseinFA,14%andspleniumofthecorpuscallosummeandecreaseinFA,16%. TheyfoundastatisticallysignicantcorrelationbetweenFAvaluesandtheseverityofheadinjury, asmeasuredwithacuteandsubacuteneurologicassessmentsacuteGCSanddischargeRankin scoresincorrelationwithregionsofobservedDAI. MacDonald,etal.conductedaDTIstudyspecicallyobservingblast-relatedtraumatic 31

PAGE 37

braininjuryinUSmilitarypopulations.Blastrelatedheadinjuryhasamuchhigherincidence ofDAIthannormaltrauma,andmaynotdisplaycognitivesymptomsexceptinthe'miserable minority'ofcasesto30%. [31] Intheirresults,18ofthe63subjectswithtraumaticbrain injuryhadabnormalitiesonDTIthatwereconsistentwithmulti-focaltraumaticaxonalinjury.This isbecauseFAwasreducedinatleasttwoormorebrainregionsforeachsubject.Abnormalities detectedonDTIweredenedasrelativeanisotropyreductionsifthevalueswereatleast2SDbelow themeanmapmadefromthe21controls.Basedontheirthresholdsforsignicanceandprojected falsepositiverates,nomorethan2of63healthysubjectscouldshow2signicantclustersby chance.MacDonaldetal.alsoobservedanadditional20subjectswithtraumaticbrain injurywhohadonlyoneabnormalitydetectedonDTIand25TBIsubjectshadnoabnormalities accordingtotheaforementioneddenitions. [106] Kumaretal.conductedastudytodeterminethecorrelationofwhitematterdiffusion changesobservedinmTBIwithdeclineofneurocognitivefunctionasmeasuredbyanumberof evaluativetestsofIQ,attention,andresponsiveness.InmoderateTBIcomparedtocontrols,they observeddecreasedFAinthespleniumofthecorpuscallosum,ageneraldecreaseofMD,increased RDinthegenuandspleniumofthecorpuscallosum,increasedRDinmidbodycorpuscallosum. anddecreasedADinthegenu.IncreasedRDinthegenuandspleniumofthecorpuscallosumwas alsoobservedformildtraumaticbraininjury.Poorneuropsychologicaloutcomeat6-monthspost injurywasassociatedwithDTabnormalitiesinthecorpuscallosum. [61] 6StatisticalAnalysisTechniquesandSoftware 6.1AlgorithmsandStatisticalMethods 6.1.1StatisticalParametricMapping StatisticalParametricMappingdescribesbothamethodologyforanalyzinganddrawinginferences aboutsetsofdatausuallyvoxels,aswellasasoftwarepackagedesignedtoapplytheseapproaches toneuroimagingdata.Firstwewilldiscussthestatisticalapproach,andgetbacktothesoftware packagelater. StatisticalparametricmapsSPMsareimagescomposedofvoxelsorpixels,whosevalues 32

PAGE 38

underthenullhypothesisaredistributedaccordingtoaknownprobabilitydensityfunction,usually theStudent'sTorFdistributions.TheseareknowncolloquiallyasT-orF-maps,respectively.The successofstatisticalparametricmappingisduelargelytothesimplicityoftheidea.Namely,one analyzeseachandeveryvoxelusinganystandardunivariatestatisticaltest.Theresultingstatistical parametersareassembledintoanimage-theSPM.SPMsareinterpretedasspatiallyextended statisticalprocessesbyreferringtotheprobabilisticbehaviorofGaussianelds.Gaussianrandom eldsmodelboththeunivariateprobabilisticcharacteristicsofaSPMandanynon-stationaryspatial covariancestructure. Statisticalparametricmappinghaschangedinitsgeneralmeaningovertime,andnowcommonlyreferstothecombinationofthegenerallinearmodelGLMandgaussianrandomeld GRFtheoryasastandardexperimentaldesignmethodologyformanipulatingdataandmakinginferencesaboutthespatialrelevanceofthevoxeldatathroughtheuseofstatisticalparametricmaps SPMs.TheGLMestimatestheparametersthatcouldexplainthedistributionofthespatiallycontinuousdata,inthesamewaythatconventionalanalysisdoesfordiscretedata.GRFtheoryisused toresolvethemultiplecomparisonproblemthatensueswhenmakinginferencesoveravolumeof thebrain.Itassumestheproducedstatisticimagetobeagoodlatticerepresentationofanunderlyingcontinuousstationaryrandomeld.Ittherebyprovidesamethodforcorrectingpvaluesfor thesearchvolumeofaSPMandplaysthesameroleforcontinuousdataasaBonferonnicorrection doesforperformingmultiplediscretestatisticaltests.Theclassicalstatisticalinferenceisusedto testhypothesesthatareexpressedintermsoftheseGLMparameters.Thishypothesistestingofthe datasetusesanimagewhosevoxelvaluesarethemselvescomputedstatisticsforthatvoxel,oruses acomputedstatisticalparametricmapT-map,Z-map,F-map,dependingonthestatistictestbeing performed. 6.1.2Voxel-BasedMorphometryAnalysis Voxel-basedmorphometryVBMisaneuroimaginganalysistechniquethatallowsinvestigation offocaldifferencesinbrainanatomyusingthestatisticalparametricmappingdescribedabove, dependentonthetypeofcomparisonbeingdone.Linearregressionsorleast-squareregressionare 33

PAGE 39

oftenusedinVBMneuroimagingstudies,comparinggroupsbysomeprimarycovariateandan additionalsetofcovariates. [73] Intraditionalmorphometryanalysis,volumeofthewholebrainorasubpartismeasuredby drawingregionsofinterestROIsonimagesfrombrainscanningandcalculatingthevolumes enclosedintheROIforeachsubject.Thisisatimeconsumingprocessandcanonlyprovide measuresoverratherlarge,smallerdifferencesinvolumebetweensubjectsmaybeoverlooked. Instead,VBMutilizesallvoxelsandassumeseachonerepresentsasimilarspatialregionbetween patientswhenperformingstatisticalcomparison.Toensurethisvoxelsimilarityassumptionismet, VBMperformspreprocessingofdataaccordingtoSPM.Firstitregisterseverybrainscan'svoxels toasimilartemplate,whichremovesmostofthelargerdifferencesinbrainanatomyshapeand sizeobservedamongstagivenpopulation.Then,thebrainimagesaresmoothedbyafunction soeachvoxelrepresentstheaverageofitselfanditsneighbors.Finally,theimagevolumesare comparedacrossbrainsateverysimilarvoxel.Parametricstatisticalmodelsareassumedateach voxel,usingtheGLMtodescribethedataintermsofexperimentalandconfoundingeffects,and residualvariability.ForfMRIcasestheGLMisusedincombinationwithatemporalconvolution model.Astatisticalparametricmapisproducedcontainingstatisticvaluesforeachvoxelcompared. Then,parametricmapvoxelsthatdon'tmeetagivenp-valuethresholdforsignicanceareexcluded. ThesurvivingclustersofSPMT-mapvoxelsarethenusedasthespatialROIstobecomparedduring furtherSPManalysis. 6.1.3IndependentComponentAnalysisICA Sinceitsemergence,dataanalysisinneuroimaginghasbeendominatedbytheuseofmass-univariate multiplelinearregressionmodels,especiallyinfunctionalmagneticresonanceimagingfMRI studies.Unlikeunivariatemethodsutilizingregressionanalysis,ICAdoesnotnaturallygeneralizeamethodsuitablefordrawinginferencesaboutgroupsofsubjects.ThismeansthatICAwill identifypotentialindependentsourcesofeffectwithinadataset,regardlessoftheiractualsignicanceorabilitytodescribetheoverallgroupanditscorrelations.TheworkingdifferenceofICA toothermethodsthatallowforcomponentidenticationisthatICAdeterminescomponentsusing 34

PAGE 40

higher-ordersignalstatisticstoproduceasetof`components'thataremaximallyindependentof eachotherbysomedesignatedfunction.Theutilityofthisisexempliedbythe'cocktailparty' scenario,whereinrecordingsofthediscussionsatapartyaretakenasdata,andthegoalistoisolate anyindividualspeakeramongstthenoisysets,suchthatyoucouldlistentojusttheirvoice. [81] Describedinmorediscreteterms,thedatatobeanalyzedisrepresentedbytherandomvector x = x 1 ;:::; x m T andtheunderlyingcomponentsastherandomvector S = s 1 ;:::; s n T .Thetaskis totransformtheobserveddatax,usingalinearstatictransformation W as S = Wx ,intomaximally independentcomponents s asmeasuredbysomefunction F s 1 ;:::; s n ofindependence. Inneuroimaging,ICAoftenusesthedifferencesinthevaluesofthedatasuchaschangesin tissuedensityordiffusivityandproximityofothervoxelsasthefunctiontomeasureindependence ofcomponents. 6.2StatisticalAnalysisSoftware 6.2.1AnalysisofFunctionalNeuroImagingAFNI AFNIisasetofCprogramsforprocessing,converting,analyzing,anddisplayingMRIandfMRI data.ItrunsonUnix+X11+Motifsystems,includingSGI,Solaris,Linux,andMacOSX.Itis availablefreeinCsourcecodeformat,andsomeprecompiledbinariesforresearchpurposes.It containsavarietyofscriptsandtoolsforperforminganalysisandapplyingvariousalgorithmsto denedvoxeldatasets. 35

PAGE 41

Figure8:AnexampleoftheAFNIgraphicaluserinterface,displayingstructuralunderlaywitha coloredT-mapoverlay. 6.2.2StatisticalParametricMappingSPM8Program ThesoftwarepackageSPMisasetofMATLABprogramsforprocessing,converting,analyzing, anddisplayingavarietyofneuroimagingdatatypesincludingMRI,fMRI,PET,EEG,andDTI. ItwasdevelopedbytheWellcomeDepartmentofImagingNeuroscienceatUniversityCollege London.ThelatestversionofthesoftwarepackageisSPM8,andfromhereonwillbereferredto assuchinordertoavoidconfusionwhendiscussingthestatisticalapproach. 7DataAcquisitionandPreprocessing Preprocessingisdenedastheapplicationofalgorithmictechniquestoalterthedatabeforeany experimentalanalysisandafterinitialacquisitionandreconstruction.Inthecontextofimaging technologies,preprocessingconsistsofcomputationalprocedurestocorrectforunwantedvariability andentropyinthedataandtopreparetheexistingdataforstatisticalanalysis.Itisthereforean 36

PAGE 42

importantstandardpracticeforallimagingstudiesandresearch.Preprocessingisdistinctfromthe experimentalanalysisbecauseitcomesbeforetheapplicationstatisticalproceduresusedtotesta hypothesis.Inthischapter,weexaminedatapreprocessingassociatedwithneuroimagingresearch. 7.1Equipment TherearetwoprimaryMRImachinesinusebytheMindResearchNetworkatthistime,a3TeslaTeldstrengthscanner,anda1.5-Tesla.5Teldstrengthscanner.BothmachinesproduceDICOMdataobjects. TherstMRmachineisaSiemens3TTriowithTotalImagingMatrixTIMApplicationSuite. ItislocatedintheMRNcomplexinAlbuquerque,NM.TheTrio's32-channelsystemiscapableof BOLDEPI,diffusion-tensorimaging,arteriolespinlabeling,perfusionanddiffusionimaging,and spectroscopy.With32usablereceiverchannelsasstandard,thesystemallowsfortheuseof4to 16current-phasedarraycoilstoimprovesensitivityandspeedofacquisition,andisreadyforfuture coildesignswithmorethan16elements.Alsoinstalled,aretheBiopacMP150DataAcquisition SystemandaBiopacGalvanicSkinResponseAmpliertoprovidereliableandconsistentreal-time monitoringofskinconductanceresponses. [100] ThesecondmachineisaSiemensMagnetomAvantosyngo1.5TMobileMRISystem.The AvantoMRmachinehasbeenmountedinsideamodiedmobileRV,tomaketherecruitmentand acquisitionprocessesmoreaccessibletolocalpopulations.TheAvantoisthemostadvancedsystem ofthisstrengthintheSiemensproductline.Itfeaturesanultra-short,150cm-longwhole-body superconductive1.5Tmagnet,with5thgenerationactive-shieldingtechnologywithcountercoils, ExternalInterferenceShieldingandexcellenthomogeneitybasedon24planplot,50cmDSV type,0.8ppm.Thesystemcomesequippedwitha12-elementMatrixheadcoilcapableofultrafastparallelacquisitionin4-CPmode,8-dualmodeor12-triplemodechannelsettings.The baseAvantosystemwasupgradedtoincludetheSQ-engineGradientSystemwithAudioComfort. TheSQ-enginegradientshavemaximumamplitudeof45mT/minuteforthelongitudinaldirection and40mT/minuteforhorizontalandverticaldirections;thegradientslewrateis200T/m/second withaminimalrisetimeof200microsecondfrom0-40mT/mamplitude.AudioComfortisan acousticnoisebufferthatleadstoa30dBattenuationofgradientnoise.Thesystemiscapableof 37

PAGE 43

BOLDEPI,diffusion-tensorimagingDTI,perfusionanddiffusionimaging,andspectroscopy. [100] 7.2FileFormats 7.2.1DICOMDigitalImagingandCommunicationsinMedicineFiles ThespreadofMRItechnologiesandthesubsequentproliferationofavailablemachinedesignsin the1970'sandearly80swassudden,andimmediatelyresearchersandmedicalinstitutionsneeded standardizedmethodsofdataprocessingforconsistencyandresearchcomparisonpurposes.DecodingtherawimagedatathatcamefromtheCTandMRImachinesofthetimewasverydifcult foranyoneexceptperhapsforthemachinemanufacturers,whoinsomecaseswantedtoenforce theirownproprietaryleformatordecodingalgorithms.Tothisend,theAmericanCollegeof RadiologyACRandtheNationalElectricalManufacturersAssociationNEMAformedacommitteetoestablishastandardformedicalimagingleswhichwasrstimplementedin1985as the'ARC/NEMA300'.Thisrstdraftwaspoorlydened,andadherencetoitsdesignamongst institutionswasirregularfortherestofthedecade,especiallyoverseas.Itwasin1993thatthe currentDICOM.dcmleformatwasestablishedasthestandardofmedicalimagingdata.The actualdetailsandprotocolsoftheDICOMleformathavebeencontinuouslyupdatedsincethen, butremainsmostlyreversecompatible.ThatistosaythatmostolderDICOMlesremainreadable andcanbeprocessedalongsidenewerDICOMformat.DICOMdiffersfromsome,butnotall,data formatsinthatitgroupsinformationintodatasets.Thatmeansthataleofachestx-rayimage,for example,actuallycontainsthepatientIDwithinthele,sothattheimagecanneverbeseparated fromthisinformationbymistake.ThisissimilartothewaythatimageformatssuchasJPEGcan alsohaveembeddedtagstoidentifyandotherwisedescribetheimage.TypicalDICOMlescontain attributesofpatientname,identicationnumberID,formattingdetails,IPandnetworktransfer protocolsandpermissions,andthepixeldataoftheimagesthemselves. 38

PAGE 44

7.2.2NeuroimagingInformaticsTechnologyInitiativeNIfTIFiles Beforetherawdatacanbeprocessed,theDICOMlesareconvertedintoaNeuroimagingInformaticsTechnologyInitiativeNIfTIle-type:'.nii'.Thisprocesstakesthepixeldataattributeof theDICOMleandrepresentsitinamatrixformatreadytobeloadedandviewedbyneuroimaginganalysissoftware,suchasSPM8orAnalysisofFunctionalNeuroImagingAFNI.Thename, patientID,voxeldimensionsandotherrelevantDICOMattributesaremadeintotheheaderforthe new.niile,whiletheattributesconcerningnetworkprotocolareremoved.NIfTiisausefulformat forchangingthemutablecharacteristicsofthedata,soastomakethemeasiertoperformgroup comparisonsandanalyze.Thematrixofpixeldataconsistsofrow,column,layer,orsometimes timescalefMRItakestheformof4Dtime-seriesdatacompressedtotina3Dmatrixasitsdimensions.Allelementsofthedatamatrixarenumericvalues,eachrepresentingasinglevoxelof theimage.InsMRIdata,thenumericentryrepresentstissuedensitywithinthevoxel;inDTIdata,it representsthetensorforthatvoxel,describingtherelativedegreesofwaterdiffusiondirectionality acrossthevoxelregion. 7.2.3HEADandBRIKFiles BRIKandHEADlesarebrainimagingdatalesgeneratedandusedbytheAFNIanalysispackage.Thelesaremulti-dimensionalpixeldata,or"voxel"datainessence,pixelswithavalue coveringaregionofthreespatialdimensions.TheAFNIformatalsosupportstime-seriesdata,or multiplevaluesateachvoxelovertime,perparticipant,orwithdifferentmeaningforexample,a betavalueandat-statistic.The.BRIKportionofthelecontainstheactualvoxeldata,whilethe .HEADlecontainsthemetadataandpatientinformationneededtointerpretthedata. 7.3Noise:Sources,Prevention,andCorrection Atthisstageinthepre-processing,theNIfTilecanbeviewedasathreedimensionalobject,but noisyaspectsofthedatastillmakeitunsuitableforanalysisorinference.Noiseisdenedusing 39

PAGE 45

Shannon'sinformationandsignaltheoryasanerrororundesiredrandomdisturbanceofauseful informationsignal,introducedbeforeorafterthedetectoranddecoder. [41] Evenunderthemostrigorousandcarefulexperimentaldesign,itisimpossibletoeliminate allsourcesofnoise.Instead,itcanbeeasiertodevelopmethodstorecognizeandthencorrector removesuchunwantedaspectsofthedata.Noisecomesinavarietyofforms:thermalnoise,system noise,andphysiologicalnoise,andeachcanbepreventedandcorrectedtosomeextent. 7.3.1Thermalnoise Heatintroducesadditionalelectronmovementinmaterials,whichcandistorttheelectricalcurrent travelingwithinthesensoroftheMRmachine.Thesedistortionsaddentropytothevaluesbeing recorded,inauniformlydistributedmanneracrossallvoxels.Thermalnoiseincreaseslinearly withtemperatureandmultiplieslinearlywithscannerstaticeldstrength3Teslaproducestwice asmuchthermalnoisevs.1.5Tesla,despitehigherresolutionpower. [20] Thus,thereisacase fordiminishingreturnsforimprovingimagingresolutionbysimplyincreasingthemagneticeld strength.Allvoxelsinascanareaffectedsimilarlybythermalnoise,independentofanatomy. Thiscanleadtogenerallynoisydataacquisition,whichcanmakescansunusableforcomparison purposes.Therefore,itiseasiesttoaccountforandpreventthermalnoisetobeginwith,ratherthan attempttondandremoveitlater.Tohelppreventthermalnoise,MRmachinesensorsarekeptat verylowoperatingtemperaturesthroughtheuseofliquidheliumcoolingsystems. 7.3.2Physiologicalnoise Subjectsthemselvescanintroducenoisetothescanacquisitionthroughthevoluntaryorinvoluntary movementoractivityoftheirbody.Thisisconsideredphysiologicalnoise.MRscansareconducted layerbylayer,imagebyimage,overtimetocomposethewholenalimage.Becauseofthis,any movementonthepartofsubjectwillcauseachangeintherelativelocationofthetissuescompared tothexedspatialpositionofthevoxelthatissupposedtomeasurethattissue. Manyformsofmovementduringscanningcancontributetophysiologicalnoise,includingau40

PAGE 46

tonomicfunctionslikeheartbeat,pulse,andbreathing,aswellasmotorfunctionslikedgetingof theneckorhead.ThisissueisexacerbatedinthecaseoffMRIacquisition,becausethescanning isalsotakenoveratime-series,increasingthepotentialforintroducingphysiologicalnoise.Other signicantsourcesofphysiologicalnoiseinfMRIincludethechangeintherateofbloodow,blood volume,anduseofoxygenovertime.Theuncontrollableaspectsofbloodmovementduringscan acquisitionaccountsforthemajorityofphysiologicalnoisethatendsupbeingintroducedduring fMRIacquisition. Variousmeasuresaretakentopreventphysiologicalnoiseduringacquisition,aswellastocorrectforanyfoundinthedataafterwards.First,subjectsheadscanbeimmobilizedeitherpartiallyor completelytopreventmovement,usuallywithaframeorwithsoftpackingcushions.Thisprevents mostrollingandtiltingoftheneck,butupwardsanddownwardsmovementcanstilloccuratrest, especiallyovertime.Subjectsarealsosometimestrainedbeforehandontheconditionstoexpect duringscanning,tohelppreventmovementcausedbydiscomfort,claustrophobia,orinresponseto loudorunexpectednoise.Priortrainingcansignicantlydecreasetheamountofmovementduring imaging,especiallywithchildrenandadolescents. Tocorrectforheadmotionthatendsupinthedatadespitepreventativeefforts,aprocesscalled coregistrationisusedtoidentifyandremovetheseartifacts.Thisprocessattemptstospatiallyalign theimagesacquiredduringasinglescansession.Whencoregisteringtheseconsecutiveimages, theywillbealignedtotasingleidealvolume.Arigidbodytransformation,whichisaspatial transformationofthevoxelsthatdoesnotchangetheshapeorsizeofthewholescan,isoften usedtoperformthecoregisteringtoanoptimalreferencevolume.Thedatavolumeismanipulated under6dimensionalparameters,toidentifyandremovethedistortioncausedbyheadmovements. Therstthreeparametersofthecorrectionalgorithmarethestretchingandshiftingofthex,y, andzaxisdimensions,whiletheremainingthreeparametersallowforrotationortwistingofroll, pitch,andyaw. [83] Anumberofhead-motioncorrectionandcoregistrationalgorithmsexist,some whichhavebeenshowntoeliminatenoisemoreeffectively,butsometimesattheriskofremoving false-positivesofheadmotionfromtheimage. [63] 41

PAGE 47

7.3.3Systemnoise Afterthermalnoise,anynoisethatisintroducedfromtheMRmachinehardwareitselfisconsidered systemnoise.Onetypeofsystemnoiseiscalledscannerdriftandcanbeasignicantcontributer tonoiseovertime,andcanruinlongitudinalcomparisonsofasubject.Typically,scannerdrift iscausedbyaslowchangeintheresonantfrequencyofthehydrogenatomsinthesample.This iscausedbythestrengthofthestaticmagneticelddriftingfromtheintendedpositionandeld strength.Evenwithconsistentpowersupplyandsuperconductingmagnets,thestrengthofthe staticeldhasthepotentialtograduallywaxorwane.Becauseofthis,thedimensionsoftheeld ofthesuperconductingmagnetchangesslightlyovertime,causingchangesinthesignalsreceived bythesensorsifleftuncalibrated. [21,20] Luckily,thescannersusedatMRNhadweeklyandmonth diagnosticsandcalibrationsdonetocontrolforscannerdrifteffects. Noisecanalsobeintroducedinthewiringofhardwarebyinterferencefromoutsidesources.In general,shieldedcablesareusedtoprotectthewiresfromunwantednoisefrequenciesinasensitive circuit.AshieldedwirecanbethoughtofasasmallFaradaycageforaspecicwireasitusesa plasticorrubbercoveringtoenclosethetruewire.Justoutsideoftherubber/plasticcoveringisa conductivemetalthatinterceptsanynoisesignal.Becausetheconductivemetalisgrounded,the noisesignalrunsstraighttogroundbeforeevergettingtothetruewire.Itisimportanttogroundthe shieldatonlyoneendtoavoidagroundloopontheshield. [20,41] 7.4Normalization,Smoothing,andSegmentation Thefollowingpre-proccessingmethodsfacilitatebetween-subjectcomparisons,thedifferentiation andselectionoutofspecictissuetypes,andreductionofnoiseartifactsduringcomparison,respectively. 42

PAGE 48

7.4.1SpatialCoregistrationNormalization Spatialnormalization,thewarpingofimagestobesttastandardtemplate,reducespotentially largedifferencesinbrainsizepotentially30%betweenanypairofsubjectsandstructureamongst subjectimages.Thisallowsforstatisticalanalysisoflargegroupsandfacilitatesanatomicalcomparisonsbetweensubjects.Howeverbiasedtemplates,subtlesexualdimorphismsofbrainanatomy, andsubsequentrelianceongrossanatomyratherthancytoarchitecturecanimpairnormalization usefulness,especiallyforclinicalsubjectevaluation.Normalizationofagroupisachievedbyregisteringeachoftheimagestothesametemplateimage,byminimizingtheresidualsumofsquared differencesbetweenthem.Therstthingtodowitheachbrainimageistomatchtheimagesto thetemplatebydeterminingthemostefcient12-parameterafnetransformation,afunctionwhich willpreservethecollinearityofpointsandtheratioofvectors,whileallowingforcomparisonto similarvoxelspaces. [83] ABayesianframeworkisused,inwhichthemaximumpriorestimateofthe spatialtransformationismadeusingexistingknowledgeofbrainsizevariabilitywithintheglobal population. [93] Steptwoaccountsforglobalnonlinearshapedifferences,whicharemodeledbya linearcombinationofsmoothspatialbasisfunctions. [84] Thenonlinearregistrationestimatesthecoefcientsofthebasisfunctionsthatminimizetheresidualsquareddifferencebetweeneachimage andthetemplate,whilemaximizingthedeformationsmoothness.Thismethodofspatialnormalizationdoesnotattempttomatcheverycorticalfeatureexactly,andonlycorrectsforglobalbrain shapedifferences.Itisimportantthatspatialnormalizationisnotperfectlyexactoroverlystringent,becausethenallthesegmentedimageswouldbetransformedtobeidenticalandnosignicant differencescouldbedetermined.Thequalityoftheregistrationmustbeashighaspossible,sothat thenormalizingtemplateimagedoesnotbiasthenalanalysis.Anidealtemplatetousewouldbe theaverageofaslargeanumberofMRimagesaspossiblethathavebeenregisteredtowithinthe accuracyofthespatialnormalizationtechnique. [83] Twoofthemostwidelyusedbraintemplatesin practicearetheTalairachspacetemplate,andtheMNIspacetemplate.TheTalairachcoordinate spacerequiresthattheanteriorandposteriorcommissureslieonastraighthorizontallineAC/PC line. [73] DistancesinTalairachspacearemeasuredwiththeanteriorcommisureintersectionasthe origin.Talairachcoordinatespaceissometimesreferredtoasstandardstereotaxicspace.TheTalairachtemplatewasoriginallyproducedfroma60yearoldFrenchwoman,whosebrainisslightly 43

PAGE 49

smallerthanaverage.Despitethispoorrepresentationofanaveragebrainforuseasatemplate, thecoordinatespaceisstillusedinpracticeasaregionalatlaswithmorerepresentativemodern templates.TheMNItemplatewasdevelopedbytheMontrealNeurologicalInstitute,toprovidea freecommontemplateforneuroimagingresearchpurposes.ThecurrentstandardMNItemplateis theICBM152,whichistheaverageof152normalMRIscansofhealthyindividualsthathaveeach beentransformedtotasimilarimagespace.TheMNIspacecoordinatesystemstillmayusea modiedTalairachatlastoinformthenamesofbrainregions. Figure9:Asagittalviewshowingtheanteriorandposteriorcommissurelinesandorientationfor Talairachspace. [111] 7.4.2MatterSegmentationGrey,White,CSF SpatiallynormalizedimagesarenextpartitionedintograymatterGM,whitematterWM,and cerebrospinaluidCSFimagemapsusingamodiedmixturemodelclusteranalysistechnique.It includesacorrectionforimageintensitynon-uniformitythatarisesformanyreasonsinMRimaging.Becausethetissueclassicationisbasedonvoxelintensities,thepartitionsderivedcouldbe confoundedbysmoothintensityvariations.Thereforeitisperformedbeforesmoothingprocedures. Segmentationisimportantforinvestigatingeffectsonacertaintypeofbrainmatter,orforrestrictingcomparisontoaknownregion.Forexample,thecorpuscallosumisprimarilywhitematter,and soanalysisofgreymatterimageswhenlookingforaneffectinthecorpuscallosumwouldlikely notproduceanyresults. 44

PAGE 50

Figure10:Shownaboveisaseriesofanatomicalcross-sectionsofamonkeybrain.Thetoprowis sMRI,followedbythethreemattersegmentations. 7.4.3SpatialFilteringSmoothing Inspatiallteringalsocalledblurorsmoothing,valuesforeachpointofdataareallowedto 'bleed'outintospatiallyadjacentpoints.Followingsmoothing,eachpointofdatarepresentsmore ofanaverageofitselfanditsneighborsthanpreviously.Smoothingcanbedonediscretelyor continuously.Forcontinuoussmoothing,thestrengthoftheeffectdecreasesinmagnitudeasone movesawayfromthedatapointuponwhichthesmoothingisbeingapplied.Themotivationsfor smoothingofdataaremultifaceted.Accordingtothematchedltertheorem,wheninvestigating aneffectthebestsmoothingkerneltobeusedisonethatdirectlyrelatestothesizeoftheeffect thatoneanticipates. [73] Iflargetumorsorhematomaarebeinginvestigated,alargersmoothing kernelrepresentativeofthateffectshouldused.GRFtheorytellsusthattherearealwayserror distributionspresentinourdata,butthattheyareadiscrete'lattice'representationofanunderlying continuousGaussianeld.Iftheerrordistributionwastrulycontinuous,wewouldexpectthe errorstonotcongregateinspecicareasorproducefalse-positiveclusters.However,sincethese 45

PAGE 51

errordistributionsarediscrete,thereexiststhechancethattheymayhavevariationsbyvoxelthat couldeffectobservedregionaleffects. [71] Bythecentrallimittheorem,smoothinganydatawill rendertheexistingerrorsmorenormalintheirdistribution.Ifwedoexpecttoseeasignicant regioninthedata,thenusingasmoothingkernelmuchlargerthanourvoxelsizewoulddisperse thediscreteerrordistributiontoastateofvirtuallyundetectableuniformity,withoutremovingthe largerandmoreconcentratedsignicantvalueofarealregionaleffect.Therefore,bysmoothing largerthanthevoxelsize,weimprovethevalidityofsubsequentinferencesmadeusingparametric testsbydecreasingtheeffectoftheerrorwithoutsacricingtoomuchoftheregionspecicity. [21] Inaddition,inthecontextofinter-subjectaveragingitisoftennecessarytosmoothmoree.g. 10mminsMRIorDTItoprojectthedataontoaspatialscalewherehomologiesinanatomyare properlyexpressedbetweensubjects.Forexample,iftwoclusterswerebothverysignicantand existedclosetooneanother,itcouldbeassumedthattheyrepresentedasimilareffectinthesame region.However,variabilitybetweensubjectsresultinginmultiplelocalmaximumswouldcause themtoappearasseparateeffectsduringregressionanalysis.Aftersmoothing,theadjacentlocal maximumswouldberepresentedtogetherbyasingleclusterwithasinglelocalmaximum.This providesabetterrepresentationoftherealeffectbeingobservedwithinthatregion. GaussianSmoothing Inneuroimaging,usuallysmoothingtechniquesapplyanisotropicornormallydistributedcontinuoussmoothingkerneloneachvoxel. [73] Thesamekernelisappliedtoeveryvoxeloftheimage. ThisiscalledaGaussiankernelbecausethesmoothingisappliedtonearbyvoxelsaccordingtoa Gaussianornormaldistribution.Whendescribingthesmoothingkernelparameterintermsofspatialdistance,wearereferringtothefullwidthathalfmaximumFWHM.Thisservesasageneral measureofthespreadofafunction,givenbythedifferencebetweenthetwoextremevaluesofthe independentvariableatwhichthedependentvariableisequaltohalfofitsmaximumvalue.see gure11TheFWHMisdependentonthevariance,andsogivesageneralideaofthemeanrange ofmaximumeffectforthesmoothingkernel. 46

PAGE 52

Figure11:Agaussiandistributionin1dimension,showingvarianceandfullwidthathalfmaximum AnisotropicSmoothing Notallsmoothingkernelsareappliedwithnormalorgaussiandistributions.Anisotropicsmoothing isusedtodescribeanyparametersforsmoothingthatarenotaGaussiankernelappliedtoallvoxels. SomeanisotropicsmoothingappliesagreaterFWHMincertaindirectionsthanothers.Otherforms ofanisotropicsmoothingwillonlyapplytheirkerneltoaspecicregionofvoxels,oruseadifferent kerneloneachvoxel,dependentonitslocation. [71] 47

PAGE 53

Figure12:Anexampleofiterativeanisotropicsmoothinganditseffectswitheachapplication.The applicationofthesmoothingkernelhasbeenrestrictedintheregionoutlinedinthesecondpanel. [71] 48

PAGE 54

PartII AStudyofStructuralMRandDiffusionTensor ImagingofmTBI 8MethodsandAnalyses 8.1StructuralMRI 8.1.1ParticipantSelectionandDataAcquisition Theoriginaldatasetconsistedofatotalof775malesubjects,396ofwhomhaveadiagnosedincidenceofTBI,and379healthycontrolsHC.Thesesubjectswererecruitedfromalocalprison bytheMindResearchNetwork.InformedconsentwasobtainedfromsubjectsaccordingtoinstitutionalguidelinesfortheUniversityofNewMexicoandalldataisprocessedandmanagedaccordingtoHIPAATheHealthInsurancePortabilityandAccountabilityActstandardsandregulations. Highresolution4-echoMPRAGET1structuralMRIscanswerecollectedbytheMRNmobileunit, whichisequippedwithaSiemensMagnetomAvantosyngo1.5Teslascanner,takingonescanper subject.Subjectsforthisanalysiswerecompensatedfortheirparticipationbyadditionalyardtime andcigarettes. 49

PAGE 55

T 1 MRIAcquisitionparametersinclude: TRrepetitiontime=2530ms 7-degreeipangle slicethickness=1.3mm FOVeldofview=256mm ImageResolution=256256 Acquisitionvoxeldimensions:1.3.0.3mm 8.1.2ParticipantInclusionCriteriaandQualityAssurance The775subjectswereobtainedbasedoninitialrequirementsthatthesubjectshadonrecord relevantTBIinformationlossofconsciousness,RivermeadPostconcussiveQuestionairreRPQ, #ofTBIandassociatedstructuralMRIscandata.OnlysubjectswithmildTBIwereeligible toparticipateinthecurrentstudy,sothosewithanyhistoryofsevereorpenetratingheadinjury wereexcluded.Similarly,subjectswithoutsufentclinicalanddemographicinformationavailable wereconsideredineligibleaswell.InclusioncriteriaforthemildTBIsamplewerebasedpartlyon theDepartmentofDefense/DepartmentofVeteransAffairsDOD/VAClinicalPracticeGuideline [94] However,lackingGlasgowComaScaleratingsforthisforensicpopulation,RPQwas usedtodetermineTBIseverityinstead.Themajorityofhealthycontrolswerethenselectedthrough acase-controldesigntomatchmTBIpatientsintermsofageandeducationeachwithin2years. Thisproduced445matchedparticipants,priortoremovalsforincompleteorpoordataqualityand undesirableneuropsychologicalconfounds. Inordertoensuretherelevanceofthedatafromeachparticipantandtoremovepotentialoutliersfromanalysis,anumberoffurtherinclusioncriteriaweredeveloped.Theremovalprocessof subjectswhodonotmeettheinclusioncriteriaforthestudywasdoneintwostages.Stage1criteria forinclusionconsistsofdemographicandclinicalconsiderations.Demographicrequirementswere thatsubjectsmustbeatleast18yearsofage,lessthan65,haveanIQgreaterthan70toexclude 50

PAGE 56

forpotentialmentalretardation,nothaveanylossofconsciousnessgreaterthan30minutesforany singleTBIincidence,andnothavepost-traumaticamnesiaoralterationofconsciousnessgreater than24hoursfollowingTBI.Clinicalrequirementwasthatnosubjectwithadiagnosedneurologicaldiseaseorpsychiatricdisorderwasincluded.AnysubjectwithhistoryofBipolarIorII,major depression,ADHD,anyschizophreniaspectrumdisorder,delusionaldisorder,psychosisNOS,or drugaddictionwereexcluded.Afterstage1inclusioncriteriaweremet,thedatabasecontained 393remainingsubjects.Stage2removesanysubjectswithpoorornoisysMRIdatathatdonot meetqualityassurancestandards,basedonaVBMcorrelationvaluelessthan0.90.Participants identiedashavingpoorMRIdataqualityn=10weresubsequentlyexcludedfromanalyses.After stage2criteriaforinclusionweremet,thedatabasecontained383subjectswithusablesMRIdata. Thisleftanaldatasetof383participantsmeanage=36.62,SD=9.95,range=18-65for analysisinthecurrentstudy,havingfullledalloftheinclusioncriteria. 8.1.3ImageProcessingandStatisticalAnalyses Anatomicalimageswerecorrectedforhead-motion,theninterpolatedtovolumeswith2mm 3 voxelsandnormalizedtoastandardstereotaxiccoordinatespace.Theimageswerethensegmented intogreyandwhitematter,andsmoothedwitha10mmFWHMGaussiankernel.Greyandwhite matterimageswereanalyzedbyvoxel-basedmorphometryVBMmethodologyusingtheAFNI softwarepackage.Theprimarystatisticaltoolusedforthiswas3dttest,anAFNIscriptimplementingagenerallinearregressionmodeloft-testevaluationon3Ddatasets.Thistestdetermines onavoxel-by-voxelbasis,iftwosetsofdataaresignicantlydifferentfromoneanother.The datasetof383participantswasseparatedintotwogroups,onecontainingallsubjectswithmTBI n=196,andtheothercontainingallhealthycontrolsn=187.3dttestwasperformedcomparing thesetwodatasets,withrespecttocovariatesincludingsubjectage,IQ,totalbrainvolume,white mattervolume,greymattervolume,totallifetimeLOC,PCLRPsychopathyevaluationmeasure, andhandedness.3dttestproducesavolumetrict-mapofthegroupcomparison,withadditional t-mapsforeachcovariate. AsecondAFNIscript,3dClustSim,wasusedtoestimatetheprobabilityofType-Ierrornoiseonlyeffectsappearingsignicantduringanalysis,AKAfalsepositiveclustersoccurringatrandom. 51

PAGE 57

Itdoesthisbyproducingmanyrandomlygeneratedimagesofthesamevoxelsizesanddimensions asthetemplateimagespaceusedinthegroupcomparison.Eachvoxelvalueisrandomlyassigned, andtenthousandfullimageswereproducedwiththeserandomvaluedistributions.Oncesimilar smoothingmethodsareapplied,theaveragenumberofsignicantlyoccurringclustersforvarious p-valueandalpha-valuethresholdsthengivesanideaofthetypeIerrorprobabilityforthosevalues. Thisprobabilityisusedtocomputeanupperthresholdofclustervolumesizerequiredtobeconsideredasignicanteffect[2x2x2x261=2088voxels].Thisthresholdaccountsfortheincreased rateoffalsepositivesovermultipletrialcomparisonsandexcludesthoseclustersizesthatcould reasonablyoccurbychance. 3dMergeandEasythresharetwomoreAFNItoolsusedtoperformsignicantclusteridentication.Theseclusteringmethodsaccountforthemultiplecomparisonrateoffalsepositivesusing theoutputfrom3dClustSimtosetthethresholdforminimumclustersize.Thesescriptsexamine theT-mapsproducedby3dttest,andwithauser-givenp-valuethresholdandthe3dClustSimcluster sizethreshold,selectoutsignicantclustersfromtheT-maps.Both3dMergeandEasythreshwere runinordertoallowforcomparisonbetweenthetwomethods.Bothresultsforidentiedsignicantclusterswereclosetoidentical.Thereforeonlythe3dMergemethodwereusedinsubsequent analysis,becauseofitsgreaterprecedenceinpreviousstudiesusingAFNI.Eachclusterobtained fromthe3dMergescriptwasconsideredseparatelyasaregionofinterest,andwerethenanalyzed acrossallsubjectsusingtheAFNItool3dROIstats;ascriptallowingforagroup-widecomparison andmeasureofeffectsizeofasingleclusterorROI. Becauseofthelargecohortsizeinvolvedinthisanalysis,thereishigherpotentialforasmallbut signicantgroupofsubjectstoskewtheclusterresults.Furtheranalysiswasperformedtodeterminetheextenttowhichtheobservedeffectsizeswerecausedbysomesignicantsubpopulation. First,totestthepossiblecontributiondifferencesofmultipleTBIvs.singleTBIsubjecteffect sizes,anANOVAwasdonecomparingsingleTBI,multipleTBI,andhealthycontrolgroupsforall covariates.Thenitwasdecidedtodetermineeffectsizesofvarioussubpopulationsnotrelatedto knowngroupingvariables.Thisprocesswastorandomlyshufeandjackknifethefulldatasetof 383subjects,andrepeatedlyselectsubpopulationsof100,with50mTBIand50healthycontrols each.100suchsub-samplingswereperformed,and3dMergeclustertestswereperformedoneach toexaminewhethersimilarclusterstothefullcohortanalysisstillappeared. 52

PAGE 58

8.2DiffusionTensorImaging 8.2.1ParticipantSelectionandDataAcquisition TheoriginalDTIdatasetconsistedofatotalof100subjects,52patientswithmildTBIand48 healthycontrols.AllpatientswererecruitedfromlocalAlbuquerqueEmergencyRoomsandwere evaluatedclinically.Healthycontrolswererecruitedthroughacase-controldesign.Specically,the majorityofcontrolsweredirectlyrecruitedtomatchpatientsintermsofgender,age,andeducationeachwithin2years.Informedconsentwasobtainedfromsubjectsaccordingtoinstitutional guidelinesfortheUniversityofNewMexicoandalldataisprocessedandmanagedaccordingto HIPAATheHealthInsurancePortabilityandAccountabilityActstandardsandregulations.DTI datawereacquiredusingatwice-refocusedspinechosequenceontheSiemens3TTriomachineat MRN,takingtwosetsofDWIscansequencespersubject.TheDTIdataconsistedofDWIimages takenalong35differentdiffusiongradients. DTIdataacquisitionparametersinclude: TRrepetitiontime=9000ms TEechotime=84ms 72axialslices,frominferiortosuperior Multi-slicemode:interleaved ImageResolution=256128 Slicethickness=2.0mm Acquisitionvoxeldimensions:2.01.02.0mm Diffusiondirections=35 B-value:800s/mm 2 53

PAGE 59

8.2.2ParticipantInclusionCriteriaandQualityAssurance The100subjectswereobtainedbasedonhavingrelevantTBIinformationLossofConsciousness,GlasgowComaScore[GCS],#ofTBIandassociatedDTIscandata.Inclusioncriteriafor themildTBIsamplewerebasedontheDepartmentofDefense/DepartmentofVeteransAffairs DOD/VAClinicalPracticeGuideline. [94] GCSof13,lossofconsciousness<30min, post-traumaticamnesialessthan24hours,alterationofconsciousnesslessthan24hours. Allparticipantsexperiencedanalterationinmentalstatus,andthemajorityofthesamplealso experiencedalossofconsciousness.MildTBIparticipantsandcontrolswereexcludediftherewas ahistoryofneurologicaldisease,psychiatricdisturbances,additionalclosed-headinjurieswith45 minorgreaterlossofconsciousness,anyheadinjurywithinthelastyear,autismspectrumdisorder, BipolarIorII,ADHD,Depression,ahistoryofsubstanceoralcoholabuse,orwereyoungerthan 18yearsold.Similarly,subjectswithoutsufentclinicalanddemographicinformationavailable wereconsideredineligibleaswell.Afterallinclusioncriteriaweremet,thenaldatasetcontained 60participantsmeanage=29.05,SD=9.85,range=19-51foranalysisinthecurrentstudy. 8.2.3ImageProcessingandStatisticalAnalysis ThetwoDTIimagedatasetswerecorrectedforheadmotion,thenconcatenatedtogetherin ordertoimproveaccuracywhenestimatingthetensors.Imageswereinterpolatedtovolumeswith 2mm 3 voxels,andnormalizedtoastandardstereotaxiccoordinatespace.Thisco-registrationused analignmentwith12degreesoffreedomandtherotationaleffectsofthisalignmentwereapplied tothegradienttable.FunctionalanisotropyFAandmeandiffusivityMDintensityimagemaps werethencomputedforeachsubjectfromtheirtensordata,andeachimagewassmoothedusinga 10mmFWHMguassiankernel.FAandMDdatawereanalyzedbyvoxel-basedmethodologyusing theAFNIsoftwarepackage.Theprimarystatisticaltoolusedforthiswas3dttest.Thedatasetof60 participantswasseparatedintotwogroups,onecontainingallsubjectswithmTBIn=30,andthe othercontainingallhealthycontrolsn=30.3dttestwasperformedcomparingthesetwogroupsfor bothFAandMD,withrespecttocovariatesincludingsubjectage,gender,andyearsofeducation. 3dClustSimwasruntodeterminetherateofTypeIerror,theoutputfromwhichwasusedtosetan appropriatesizethresholdforsignicantclusters.3dMergewasruntoselectforsignicantclusters 54

PAGE 60

intheFAandMDimages.Eachclusterobtainedfromthe3dMergescriptwasconsideredseparately asaregionofinterest,andwerethenanalyzedacrossallsubjectsusingtheAFNItool3dROIstats; ascriptallowingforagroup-widecomparisonandmeasureofeffectsizeofasingleclusterorROI. 55

PAGE 61

9Results Whenobservingtrendsindataandattemptingtomakemeaningfulinferencesaboutthepopulationthatthedatarepresents,itisimportanttoconsiderthemeaningofcorrelationsbetweenas manyfactorsandcovariatesaspossible,notjusttheprinciplecorrelationbeinginvestigated.The variouscorrelationsthatexistwithintheobservedcohortrevealpotentialdifferencesfromtheglobal population. 9.1StructuralMRIResults Whiteandgreymatterimageanalyseswereperformedidentically.AnalysisofT 1 -weighted whitematterstructuralimagesproducednosignicantlysizedclustersbetweenmTBIgroupand healthycontrols,butnumeroussmallclustersoftissuedensityabnormalitiesexistedthroughoutthe prefrontalcortex.Althoughtheseclustersweretoosmalltobeconsideredsignicantindividually, theirpatternofexpressionalongtheseparationofwhiteandgreymattertissuesappearsindicative ofshearinganddiffuseaxonalinjuryDAItypicalofmTBI.Thismayrepresentarealdetectionof mTBIpathologyatasub-thresholdlevel,consistentwiththeheterogeneouspatternofdamagepresentation.Inaddition,abelowthresholddecreaseinwhitemattertissuedensityexistedwidespread throughoutthecerebellum.Becauseofthelackofsignicantwhitematterclusters,theremaining sMRIanalyseswillonlyobservegreymatterimages. 9.1.1NeuropsychologicalandClinicalMeasures: TherewerenosignicantdifferencesbetweenmTBIandhealthycontrolgroupsp>.05onmajorneuropsychologicalvariablesofIQ,VBMcorrelationvalue,orforhandpreferenceasassessed bytheEdinburghHandednessInventoryOldeld1971.Nosignicanttrendexistedbetween mTBIandhealthycontrolgroupsforglobalparenchymalvolume.Nosignicantdifferenceswith healthycontrolsexistedbasedonmultipleorsingleTBIgroupingsp>.05foranyvariableduring ANOVAtesting.Thismeansthegroupswererelativelysimilarintermsparticipantcomposition, andthatthematchedsubjectdesignwaseffective. SignicanttrendsbetweenthemTBIandHCgroupswerefoundhowever,forvariablesofage 56

PAGE 62

andfactor2ofthePCLRinformation,foreachsubject.Age:t=1.904,p=0.058,df=381|PCL Factor2:t=-2.038,p<0.05,df=356Thisimpliesthatonaverage,themTBIgroupwereslightly older,andthattheHCgrouphadhigherpsychopathyratingsononemeaureofthePCLR.Sinceage isastrongcorrelatewiththechanceofhavingadiagnosedTBI,itisunderstandablethatthemTBI grouphadaslightlyhigheraverageage.ThehigheraveragePCLRratinginHConlyexistedforone factorofthewholeinventory,andduetothehigherrateofpsychopathyinprisondemographics,it isreasonabletoassumethistrendoccurredbychance. 9.1.2DemographicCorrelations: APearson'srcorrelationcoefcientwasobtainedforallpair-wisecomparisonsofavailable cohortdemographicvariables.Therewasasignicantpositivecorrelationbetweentotallossof consciousnessandthemeanvalueoftheclusterrelatedtothegroupdifferentiation.r=0.145,p< 0.05.Ifanr 2 istaken,thiseffectsizebecomesalmostnegligible,howeveritisworthmentioning becausethesamecorrelationhasbeennotedinpreviousliteratureatgreaterstrength. Figure13:LogarithmicscaleoflossofconsciousnessontheYaxis,plottedagainstthemeantissue densitywithintheprimaryTBI/HCclusterontheXaxis.ThePearson'srforthedataisshownby theline. Therewasasignicantcorrelationbetweensubjectageandwholebraingreymattervolumer= -0.492,p<0.001,ageandcerebrospinaluidvolumer=0.149,p<0.05,ageandvbmcorrelation r=-0.180,p<0.05,IQandgreymattervolumer=0.175,p<0.05,IQandwhitemattervolume 57

PAGE 63

r=0.190,p<0.01,andhandednessandIQr=-0.239,p<0.01. 58

PAGE 64

9.1.3SignicantClusterRegionsofInterest: Atp=0.001,nosurvivingclustersexistedafterremovalofclustersbelowthresholdsizes.All clustersinsMRIanalysisshownforp=0.005.Theclustert-statsrangefrom-1to1representing theeffectsizeofincreaseordecreaseingreymatterdensity,respectively. t-stat+/GMDecrease+t GMIncrease-t 0 : 005 < t < 0 : 33 Red DarkBlue 0 : 33 < t < 0 : 66 Orange Blue 0 : 66 < t < 1 : 00 Yellow LightBlue Table1:ClusterSignicanceColorKey Healthy Control vs mTBI Figure14:PrimaryclusterdifferencebetweenHCandmTBIgroups,showsadecreaseingreymattervolumeintherighttemporallobe.Axial,Sagittalviewsofsignicantcolorclustersoverlayed ontoastructuralbraintemplate.RenderedinAFNI 59

PAGE 65

IQ Figure15:LowerIQcorrelatedwithdecreasedgreymatter,andtoalesserextent,whitematter volumeinthecerebellum.Axial,Sagittalviewsofsignicantcolorclustersoverlayedontoa structuralbraintemplate.RenderedinAFNI Total Volume Figure16:Greatestdifferencesintotalbrainvolumewerecorrelatedwithdecreasedgreymatterobservedinthesuperomedialcingulatecortexborderingwhite/greymatterseparation.Axial,Sagittal viewsofsignicantcolorclustersoverlayedontoastructuralbraintemplate.RenderedinAFNI 60

PAGE 66

9.2DiffusionTensorImagingResults FAandMDmapanalysiswereperformedidentically.TheresultantsignicantclustersandassociatedregionsofinterestareobservedforFArst,followedbyMD. 9.2.1NeuropsychologicalandClinicalMeasures: TherewerenosignicantdifferencesbetweenmTBIandhealthycontrolgroupsp>.05onmajor clinicalvariablesofgender,age,oreducation.Thisassuresthatthegroupswererelativelysimilar intermsofparticipantcomposition,andthatthematchedsubjectdesignwaseffective. 9.2.2DemographicCorrelations: Duetohavingonlyage,education,andgenderascovariates,demographicanalysisfortheDTI studyismorelimitedthanthesMRI.APearson'srcorrelationcoefcientwasobtainedforallpairwisecomparisonsofavailablecohortdemographicvariables.Asmalltrendexistedbetweenage andgenderr=0.128,p<0.05,indicatingthatonaverage,malesinthiscohortareslightlyolder thanfemaleparticipants.Therewasaveryweakcorrelationbetweenageandeducationr=0.0719, p<0.05,suchthatoldersubjectsweremorelikelytoholdhighereducationdegrees.Thiswas expectedbecausesomeoftheyoungersubjectswerestillenrolledandcouldnothaveachieved additionalyearsofeducationyet.Acorrelationalsoexistedbetweengenderandeducationr= -0.345,p<0.05.Femaleparticipantsweremorelikelytoholdhighereducationaldegrees. 9.2.3SignicantFractionalAnisotropyClusterRegionsofInterest: AllclustersinFAanalysisshownforp=0.001.Thefollowingclustert-statsrangefrom-1to1, representinganincreaseordecreaseinfunctionalanisotropy,respectively. 61

PAGE 67

t-stat+/FADecrease+t FAIncrease-t 0 : 001 < t < 0 : 33 Red DarkBlue 0 : 33 < t < 0 : 66 Orange Blue 0 : 66 < t < 1 : 00 Yellow LightBlue Table2:ClusterSignicanceColorKey FA change in HC vs. mTBI Figure17:AdecreaseinFAisobservedinthebodyofthecorpuscallosum,aswellasasmaller decreasealongtheanteriorbrainstemandmedullaoblongata.Axial,Sagittalviewsofsignicant colorclustersoverlayedontoastructuralbraintemplate.RenderedinAFNI Figure18:TOP:descendingaxialmontageoftheFAdecreaseincorpuscallosumcorpuscallosum,BOTTOM:advancingsagittalmontage,revealinganincreaseinFAintheinterhemispheric spaceabovethecorpuscallosum,anddecreasedFAintheanteroinferiorregionofthecorticospinal tractneartheforamenmagnum. 62

PAGE 68

HC vs. mTBI FA change differing by gender Figure19:GreatermagnitudeFAincreaseinmalesseeninthesuperiorcerebellarpeduncleAKA theCorporaQuadrigemina,includingtheinferiorandsuperiorcolliculi.Axial,Sagittalviewsof signicantcolorclustersoverlayedontoastructuralbraintemplate.RenderedinAFNI Figure20:ShowinggreaterincreaseinFAformTBImalescomparedtomTBIfemales.TOP: descendingaxialmontageofcluster.BOTTOM:advancingsagittalmontageofcluster 9.2.4SignicantMeanDiffusivityClusterRegionsofInterest: AllclustersinMDanalysisshownforp=0.001.Thefollowingclustert-statsrangefrom-1to 1,representinganincreaseordecreaseinmeandiffusivity,respectively. 63

PAGE 69

t-stat+/MDDecrease+t MDIncrease-t 0 : 001 < t < 0 : 33 Red DarkBlue 0 : 33 < t < 0 : 66 Orange Blue 0 : 66 < t < 1 : 00 Yellow LightBlue Table3:ClusterSignicanceColorMap MD change in HC vs. mTBI Figure21:AnmildincreaseinMDisobservedinthebodyofthecorpuscallosum,aswellas anincreasealongthemedialprefrontalcortex.Axial,Sagittalviewsofsignicantcolorclusters overlayedontoastructuralbraintemplate.RenderedinAFNI Figure22:TOP:ascendingaxialmontageoftheMDincreasealongthecorpuscallosum,internal capsule,andprefrontalcortex,BOTTOM:advancingsagittalmontage,revealingtheincreasedareas ofMDlateraltothecorpuscallosum,andinthemedialprefrontalcortex 64

PAGE 70

HC vs. mTBI MD change differing by gender Figure23:GreaterdecreaseinMDinmalesseeninthesuperiorcerebellarpeduncleAKAthe CorporaQuadrigemina,includingtheinferiorandsuperiorcolliculi.Axial,Sagittalviewsof signicantcolorclustersoverlayedontoastructuralbraintemplate.RenderedinAFNI Figure24:ShowinggreaterdecreaseinMDformTBImalescomparedtomTBIfemales.TOP: descendingaxialmontageofcluster.BOTTOM:advancingsagittalmontageofcluster 65

PAGE 71

HC vs. mTBI MD change differing by education Figure25:AnincreaseinMDcorrelatedwithmoreyearsofeducationobservedinthegenuof thecorpuscallosumSagittalviewofsignicantcolorclustersoverlayedontoastructuralbrain template.RenderedinAFNI Figure26:Advancingsagittalmontage.ShowsMDincreaseinthegenucorrelatedwitheducation. 10Discussion FirstwewillexaminethesignicanceofthesMRIresults.Onlythreesignicantclusterswere identiedatthep=0.005threshold.ThestructuralMRIdetectedasignicantregionofdecreased greymatterintherighttemporallobeofthemTBIsample.Thisregionwastheprimarycluster differentiatingbetweenthemTBIpatientsandthehealthycontrolsubjects.Thepositionofthis clusterwithinthemiddleofthehemispheremaybeinuencedbythecomparisonofsuchlarge samplesizes.Regionsofindividualsubjectdifferencesaremorelikelytooverlaptowardsthe centerofthehemisphere.Theclusteralsoliesdirectlyunderthethinpterionskullregion,which isatriskforsustainingdamagetotherelativelyweakmiddlemeningealartery.Acorrelationwas foundbetweenthemeanvalueofvoxelswithintheprimarycluster,andasubject'stotallifetime lossofconsciousnessLOC.Thisconrmsexistingliteraturethathasfoundstrongeroutcome 66

PAGE 72

correlationswithLOC,andsupportsthediagnosticuseofLOCasapredictivemeasureofinjury severity. [94,91] Ipsilateralexpressionofinjuryismuchmorecommonthancontralateralexpression counter-coupinjury.Therefore,thepositionofthisclusterwithinthesubcorticalgreymatterof therighthemispheremayalsobeduetosubjectinjuriesbeingmorecommonontherightsidewithin theobservedsample. ThegreatestchangesinIQwerecorrelatedwithgreyandsomewhitematterdecreaseinthe cerebellum,aregionwhichisimplicatedwithmotormemoryandattentionalcoordination.This correlationofcerebellargreymattervolumewithgeneralintelligenceasmeasuredbyIQorsimilar inventorieshasbeenshownpreviously. [107] However,allofthesubjectswithheadinjuryhadtheir IQtestsdonepostinjury.Therefore,itispossiblethatdecreaseofgreymatterinthecerebellum correlatesnotinanactualdecreaseinintelligence,butsimplywithanimpairedabilitytoanswer questionsintendedtomeasureintelligence. Theclustercorrelatedwiththegreatestchangeinoverallbrainvolumewasinthesuperomedial rightcingulategyrus,lateraltotheinterhemisphericssure.Thisappearstobeaclearindicationof lateraltraumaofgreaterseveritybeinglikelytocausecingulateherniation,leadingtocompression againstthefalxcerebri,theinterhemisphericseparatingtissueoftheduramater.Cingulateherniationexhibitshighriskforfurtherischemiaandtissuenecrosis,leadingtotheobservedcorrelation forlossinoverallbrainvolume.Thisexpressionofinjuryisalsoconsistentwiththeright-sided preferenceofinjurydisplayedwithinthecohort.Thegreaterseverityofanyinjuryalsousually resultsinincreasedlossofvolumeofalltissues.DespitetheseinterestingresultsseeninthesMRI studysforensicpopulation,thereremainanumberofconfoundstoconsider.Thepopulationobservedisquitelargewith383subjects,anddespiteclustersimulationtoaccountfortherateofType Ierror,theriskisalwayspresentwithlargersizesamples.Jackkningtrialsof100subjectsubpopulationsonlyproducedsignicantresultsoverlappingtheoriginalclusterregionsin30-46%of samples.Thisshowsthatwithinthesample,theeffectsizesoftheseobservedclustersismoderateat best,andmaynotbestronglyrepresentativeofthegeneralpopulation.Inaddition,apotentialconfoundoftheforensicpopulationisahigherrateofco-morbidityofotherdiseasesandneurological disorders,whichmayhaveeffectsdespiteeffortstoexcludediagnosedsubjectsfromparticipation. Theuseoftherivermeadpost-concussivequestionnaireasaselfreportisnotasstronglyconrmed forseverityofinjuryastheGlasgowcomascale,butremainsadecentpredictivemeasureofhead 67

PAGE 73

injuryincludingmTBI. ThereremainspoorconsensusinsMRIresearchregardingthebrainregionsinwhichmTBImay haveaconsistentexpressionofdamage.OursMRIresultsindicatenoregionsstronglyimplicated byprevioussimilarmTBIresearch.ThisisinpartduetotheheterogeneouspatternofinjuryexpressionformTBI,butisalsolikelyduetothepoorabilityofsMRItodetectdiffuseaxonalinjury, asevidencedbytheweakeffectsizeofthejackkningreproducibility.Thecorpuscallosum,and especiallythegenu,isoneoftheonlyregionswhichhasconsistentlyexhibitedatrendofobservable changesinmTBIresearch.Thisislikelybecausetheregionhasalargeconcentrationofunmyelinatedneurons,whicharemoresusceptibletotheDAIfromtorsion,shearingorstretchingontheir unprotectedaxons. [65] Finally,theexpressionofwhitematterabnormalitiesatsub-thresholdlevels dispersedthroughouttheprefrontalcortexappearconsistentwithmulti-focalDAI,butnosound inferencescanbemaderegardingtheirsignicance. Thediffusiontensorimagingresultsrevealedsignicantclustersforthemajorityofdemographicvariablesavailableatap=0.001threshold.Asexpected,theprimaryclustersdifferentiating themTBIandhealthycontrolgroupsshoweddecreasedFAandincreasedMDinthebodyofthe corpuscallosum,whiletheincreaseinMDextendstotheinternalcapsuleaswell. [69,64,57] Based onrepeateddetectionofsignicantregionsinpreviousDTIstudies,thethreeareasmostcommonly involvedinDAIarethesubcorticalwhitematter,corpuscallosum,andthedorsolateralaspectofthe upperbrainstem. [15] Additionally,FAandMDarefrequentlyobservedasbeinginverselyrelated, regardlessofregion. [15] Thecorpuscallosumisatgreaterrisksincecloseto80%ofitsbersare unmyelinated,whichexperiencetraumaticaxonalshearingfromlessforcethanmyelinatedbers. Whenlateralaccelerationordecelerationoccurs,themoststrainisputontheregionsconnecting tissuesofdifferentdensities,becauseeachtissuereactsatdifferentspeeds.Becausethecorpus callosumisthemajorinterhemishpericcommisure,italsoexperiencesmorestrainfromlateral separationofthehemispheresthanmostotherbrainregions. [70,65] Primaryclustersalsoincluded decreasedFAintheanteroinferiorcorticospinaltractalongtheforamenmagnum,andincreasedFA inanareawithintheinterhemishpericssuresuperiortothecorpuscallosum.Thecombinationof greatlydecreasedFAandincreasedMDalongthebodyofthecorpuscallosumindicatesdisruption ofaxonalcytoskeletonmembranes.Theshearingandstretchingonthisregionleadstobreakdown ofthecytoskeleton,viaCa 2+ mediateddegeneration. [99,44] Thiscausesincreasedpermeabilityof 68

PAGE 74

theaxolemmaandeventuallyapoptosis.Bothleadtoanincreasedamountofextra-cellularspace, resultinginthegreaterMDobservedwithinthearea.Suchdamageistypicalofwhitemattertract injuryinmTBI,andsupportspreviousndingsofreducedFAandincreasedMDinthecorpus callosum. [25,27,33,56] TheincreasedFAintheinterhemisphericssureabovethecorpuscallosumseemsunusualat rst.However,becausetheexpectedFAforthisregionisverylowalready,itisverylikelythatother moreanisotropictissuesnowoccupythatsamespaceinmTBIsubjects.Thisdisplacementcould befrommovementcausedbytheprimaryinjury,orbyhematomaandcytotoxicswellingleadingto tissuesbeingpushedintothatarea.Thepossiblepresenceofadisplacinghematomaissupported bythedecreasedFAinthebrainstemandcorticospinaltract,indicativeofradialcompressionalong thespinalopeningtotheskull.Asignicantclusterwasobservedbetweengenderdifferencesin theexpressionofFAandMDchangesforhealthycontrolsandmTBIpatients.Thesuperiorand middlecerebellarpeduncles,tractsofwhitematterbersconnectingtothecerebellumandcontainingthesuperiorandinferiorcolliculi,experiencedincreasedFAanddecreasedMDindicativeof acute-phasecytotoxicedema.Thisacute-phasecellularswellingwithintheareaoftenoccursfrom forwardwhiplash.TheresultsindicatethatmTBImaleswithinourcohortexperiencedgreaterFA andMDchangeswithinthisregionthantheirfemalecounterparts.Thisindicationofincreased cytotoxicedemaincohortmalesmaysimplyrepresentabiasforforwardwhiplashinjurieswithin thestudysmalesubjectpopulation.Alternatively,theresultcouldrepresentarealdifferenceinthe riskofmalestoexperiencecytotoxicedemainthesetissues.Thiswouldbesupportedifsignicant whitematterstructuraldifferencesbetweenmalesandfemalesexistwithinthesuperiorandmiddlecerebellarpeduncles.Asignicantclustercorrelatedwithyearsofeducationwasobservedin thegenuofthecorpuscallosum.AgeneralincreaseinMDwithintheclusterwasassociatedwith mTBIsubjectswhohadreceivedmoreyearseducation.AbnormalitieswithinthegenuofthecorpuscallosumisconsistentlyimplicatedinmTBIinjury,andiscorrelatedwithpersistentcognitive impairment. [64] Becausethisclustercorrelatedwithincreasedyearsofeducationdoesnotcoincide withincreasesinage,wecanruleoutageasacausalfactorforthisincreasedrisk.Moreyears educationandlearningcouldleadtogreaterwhitematterbundleconnectivity,synapticstrength, andvolumewithinthecorpuscallosum,tofacilitatethegreaterdemandsofinterhemishpericcommunications.Thiswouldputthesesubjectsatriskforgreaterchangesindiffusioncharacteristics 69

PAGE 75

withinthearea.AnincreaseinMDwithnoassociatedchangetoFAwithintheregioncouldimply eitherminordiffuseaxonalinjurywithoutadisruptionofoverallwhitemattertractintegrity,or theadditionalincreasedpresenceofreactivemicroglia.Increasedmicrogliapresenceinresponse toinammationandtissuedamagewouldincreaseMDwithinthatregionwithoutachangeinFA, andwouldslowlydecreasebacktobaselineMDlevelwithinafewweeks.Thisisbecausemicrogliaareblobbymacrophages,andwouldhaverelativelyisotropicdiffusioncharacteristics.This addeddiffusivitywouldshowuplessandlessastheregionrecoveredandexhibitedlowermicroglia counts.Afollow-upDTIcheckingforMDdecreasewithinthegenucouldconrmthemicroglia hypothesis. 11Conclusion Overall,theDTIstudyproducedgreaternumbersofsignicantclusters,andatahigherthreshold forsignicancethantheresultsofthesMRIstudy.ThesMRIstudypresentedanumberofpotential confounds,andrelativelyweakeffectsizesforsurvivingclusters.ThissMRIstudysupportsthe useoflossofconsciousnessasapredictorofinjuryseverity,theimplicationofcerebellarinjuryfor cognitiveimpairment,andthecommonlateralbiasofheadinjuries.Fromtheseresults,however,T1 structuralimagingdoesnotappearabletoshowconsistentdifferencesofmTBIfromhealthycontrol brainsinawaythatwouldbeclinicallyusefulfordiagnosis.Especiallyinlargepopulations,the effectsizesofunknownvariancesinsubpopulationsobscurethemTBIstructuralchangeswewish toidentify.TheabilitytodetectDAIandsimilarsmallchangesinphysiologyfollowingmTBIis imperativetoaccuratediagnosisoftheinjury,andsMRIseemscapableonlyofdetectingmoregross anatomicalabnormalities,suchashematomaandlesion.LikeCTscan,sMRIremainsimportantin observinglarge-scalephysicaldeformitiesoftenoccurringinmoresevereTBI. Comparatively,resultsfromtheDTIstudyhadmuchlargereffectsizesandproducedmoresignicantmTBIaffectedregionsofinterest.EachobservedclusterinDTIwasmoreinformativeas tothenatureoftheunderlyingpathophysiologythanthoseseeninthesMRIstudy.ThisDTIstudy supportspreviousndingsofthecorpuscallosumasaregionofinterestinmTBI,inlightoftheotherwiseheterogeneouspatternofinjury.ItalsosuggeststhatDTIiseffectiveatdetectinghematoma 70

PAGE 76

orotheruidbuildup,animportantfactorforheadinjurymortalityandshort-termtreatment.A clearregionofcytotoxicedemawasalsodetected,conrmingDTIefcacyfordetectingcellular swellingofwhitematterberbundles. [66] Thistypeofswellingmaycorrelatewithadelayedonset ofcognitiveimpairment,andmaybeanimportantbiomarkerinacutemTBIconsideringthatmany patientsdonotdevelopbehavioralsymptomsuntilaftermedicalattentionorimaging.DTIhas provenusefulinidentifyingregionsatriskforDAIandothershearingtypedamage,butDTImay alsobesensitivetochangesbesideshistologicalorstructuralones,suchastheincreasedpresenceof reactivemicroglia. [70,61] RegardlessofwhethertheobservedchangesimplyadetectionofDAIin thesubjectorsomethingelse,theabilityforthemodalitytoconsistentlyidentifysignicantchange inmTBIgivesitgreatpotentialasapredictivemeasurefordiagnosis. Therefore,basedonthesendingsandthesupportingresearch,achangetocurrentTBIdiagnosticandimagingprotocolisencouraged.CurrentprocedureforheadinjuryperformsCTscanasthe initialcheck.IftheCTscanpresentsnegativeforpathologies,thepsychologicalandcognitiveevaluationisusuallyusedtodeterminetheappropriatefollow-up.Ifnomajorcognitiveimpairmentsare observed,thepatientisoftenreleasedwithoutfurtherfollow-up. [10] Ifcognitiveimpairmentsuchas <15ontheGCSorothersymptomsarepresent,astructuralMRIisthedefaultimagingfollow-up. Oursuggestionistoexpandthecriteriaforsuggestedfollow-up,andthatfollow-upbeperformed usingDTIinsteadofsMRIasthestandardforevaluation.Expandedcriteriawouldincludeallcurrentmeasures,pluspresentationofanypost-traumaticamnesiaorlossofconsciousnessfollowing trauma.SinceLOCandPTAaresucheffectiveindicatorsforseverityofinjury,theirmerepresence warrantsfollow-upimaging.Unfortunately,itisnotreasonabletoassumethatthemajorityofclinicalsettingswouldbeabletoaccommodatethesuggestedprotocolchangestoheadinjurydiagnosis. DTIcanbecostly,takealongtimetoperform,andevenlongertomanuallycomputeintomapsof diffusivitymeasures.NotallMRImachinescanbeconguredtoscandiffusionweightedimages either,whichpreventsthemfromproducingDTIdata.However,asthetechnologyimprovesand becomesmorewidelyavailableandeconomicallyfeasible,thecurrentrestrictionsdisappear,and diagnosticpracticesshouldalwaysupdatetoproducethebestpossibleoutcomesforpatients.The suggestionremainsthatDTIshouldbemorewidelyusedinclinicalfollow-upimagingofhead injuries,toimprovediagnosticrecognitionofmildtraumaticbraininjury. 71

PAGE 77

Futurework Thestudiesconductedinthisthesisarelimitedbytheirseparatesubjectpopulations,preventing directcomparisonofresultsexceptforgeneralobservationsofstatisticalsignicance.Therefore, futureresearchshouldperformanalysisusingsMRIandDTIdataofthesamepopulation.TheDTI studysubjectpopulationalreadyhavestructuralMRIanatomicaldatathatcouldbeobtainedfrom MRNandprocessed,andonwhichsimilarvoxel-basedmorphometryanalysiscouldbeperformed. ThiswouldallowfordirectcomparisonofobservableclustersinsMRIandDTIscansforeach subject.Thewithin-subjectcomparisondesigncouldeasilybeaddedtothebetween-subjectgroup analysisalreadydone,andwouldprovideimportantinsightsintowhichaspectsofthesameinjury aredetectablebyeachmodality.Secondly,theDTIanalysiscouldbeimprovedbyaddingadditionalcovariatessuchassubjectGM,WM,andCSFvolumes,andotherdemographicinfoforeach participant.SimilarresearchstudieshavealsoproducedsignicantresultsusingothertypesofDTI measures,suchasbrainmapsofapparentdiffusioncoefcient,radialdiffusivity,andtractographybasedspatialstatistics. [61,68,67] Followupresearchtothisstudycouldcomputeandanalyzethe groupdifferencesforthesemeasuresaswell. Futuredirectionsofresearchbeyondthepresentstudiesshouldseektheestablishmentofdiagnosticprotocolsthatincludemulti-modalimagingtocharacterizebrainalterationsinmTBI,as oneimagingmodalitymaynotaccuratelycapturebrainalterationsinmTBI.Thereremainsgreat potentialforlongitudinalstudiesofmilitarypersonnelandcombatsettingprevalenceofheadinjury. Greatresearchopportunitieswithnumerouspracticalapplicationslikeblast-relatedDAIprevention andotherdeliverablesmaybepossible.TheUSremainsinapositionwherenewneuroimaging studiesconductedincooperationwiththeArmyandtheDepartmentofDefence/VeteransAffairs couldgreatlybenetcombatsettingtreatments,andlong-termrecovery.Entireplatoonsofsoldiers couldreceievehealthysMRI,DTI,andfMRIorMRSimagingdonepriortotheirdeployment.The abilitytoworkwithapopulationwhohaveallhadanextensivebatteryofhealthydataalreadyon recordisinvaluable,notonlyfortheextensiveandpristinepotentialsforresearchdesign,butfor thepersonalizedtreatmentoptionsparticipantswouldbeabletoreceiveinreturn. FurtherresearchforimprovingneuroimagingdiagnosticsshouldaimtodevelopgreaterautomationofmethodstoacquireandcomputeDTIimagemapsofFA,MDandotherdiffusivitymeasures. 72

PAGE 78

Thegreatereaseofuseindataacquisitioncouldsignicantlyimprovetheavailabilityanduseof DTImodalitiesinclinicalsettings.Withimprovedautomationofdataacquisition,preprocessing, andcuttingedgeapplicationsofmachinelearningclassierstopredictdiseasestates,theemergence ofalmostfullyautomatedneuroimagingdiagnosticscouldoccurwithinthenextfewdecades.Imagineifasubjectcouldgetabrainscandone,andallpreprocessingandanalysisiscomputedbefore thepersonleaves.Abatteryofneurologicaldisorderandinjurytrainedmachinelearningclassiers couldserveasanearlywarningsystemdirectingmoreinvolvedhumaninvestigationandanalysis.Similarmachinelearningclassierscouldalsouseinitialsymptomatologyreportstoprescribe theappropriatemulti-modalfollowupscansthatwouldprovidethemostinformationformaking accuratediagnoses.Itisimportantthatfutureemergingautomateddiagnostictechnologiesnotbe usedassubstituteforrealclinicalevaluationbymedicalprofessionalsandtechnicians,butinstead tosupplementthemandmakeneuroimagingdiagnosticadvancementsmorewidelyavailabletothe population. Appendix AFNILicensingandCopyright AFNI,itsassociatedprograms,anditsdocumentationareprovidedasis,andnowarrantyfortheir correctnessorusefulnessforanypurposeismadeorimpliedbytheMedicalCollegeofWisconsinMCW,theNationalInstitutesofHealthNIH,bytheauthorsofthesoftware,orbyanyone else.NeitherMCW,theNIH,northeauthoracceptanyliabilityforanydefectsinthissoftware oritsmanuals,orforanydamagescausedbyuseofthissoftware.Clinicalapplicationsarenot recommendedoradvised;thesoftwareisdesignedforresearchpurposesonly. AFNIisdistributedfreelyundertheGnuGeneralPublicLicense.MajorportionsofthissoftwarewerewrittenattheMedicalCollegeofWisconsin,whichownsthecopyrighttothatcode.For completedetails,seethelehttp://afni.nimh.nih.gov/pub/dist/src/README.copyright. 73

PAGE 79

References [1]TuongH.Le&AlisaD.Gean.NeuroimagingofTraumaticBrain Injury,PublishedonlineinWileyInterSciencewww.interscience.wiley.com. DOI:10.1002/msj.20102 [2]AnnC.McKee,etal.ThespectrumofdiseaseinchronictraumaticencephalopathyBraindoi:10.1093/brain/aws307 [3]BoakeC,DillerL.Ahistoryofcognitiverehabilitationofbrain-injuredpatients,1915.JournalofHeadTraumaRehabilitation1989;4:1 [4]JoshuaE.Wilk,etal.MildTraumaticBrainInjuryConcussion,PosttraumaticStressDisorder,andDepressioninU.S.SoldiersInvolvedinCombatDeployments:AssociationWithPostdeploymentSymptoms.:Psychosom MedApril201274:249-257;publishedaheadofprintFebruary24,2012, doi:10.1097/PSY.0b013e318244c604 [5]JenniferH.Olson-Madden,etal.IdenticationandTreatmentofTBIandCooccurringPsychiatricSymptomsAmongOEF/OIF/ONDVeteransSeekingMental HealthServicesWithintheStateofColorado:EstablishingConsensusforBestPractices.CommunityMentalHealthJournal-10.1007/s10597-012-9572-4 [6]Bogner,J.A.,&Corrigan,J.D..ReliabilityandvalidityoftheOSUTBI identicationmethodwithprisoners.JournalofHeadTraumaRehabilitation,24, 279 [7]CifuDX,CohenSI,LewHL,JaffeeM,SigfordBThehistoryandevolution oftraumaticbraininjuryrehabilitationinmilitaryservicemembersandveterans.Am JPhysMedRehabil;89:688. [8]Alexander,MichaelP.Mildtraumaticbraininjury:Pathophysiology,natural history,andclinicalmanagement.Neurology,Vol45,Jul1995,1253-1260. [9]LevinHS,BentonAL,GrossmanR."Historicalreviewofheadinjury".NeurobehavioralConsequencesofClosedHeadInjury.Oxford[Oxfordshire]:Oxford UniversityPress.pp.3.ISBN0-19-503008-7. [10]GranacherRA.TraumaticBrainInjury:MethodsforClinical&ForensicNeuropsychiatricAssessment,SecondEdition.BocaRaton:CRC.ISBN0-8493-8138-X. [11]ScurlockJA,AndersenBR.DiagnosesinAssyrianandBabylonianMedicine: AncientSources,Translations,andModernMedicalAnalyses.Urbana:Universityof IllinoisPress.p.307.ISBN0-252-02956-9.Retrieved2008-11-08. [12]SanchezGM,BurridgeAL."DecisionmakinginheadinjurymanagementintheEdwinSmithPapyrus".NeurosurgicalFocus23:E5. doi:10.3171/foc.2007.23.1.5.PMC17961064.PMID17961064. [13]NadaAndelic,Theepidemiologyoftraumaticbraininjury.Neurology: November22,2012http://dx.doi.org/10.1016/S1474-4422-6 74

PAGE 80

[14]Ermer,E.,Cope,L.M.,Nyalakanti,P.K.,Calhoun,V.D.,&Kiehl,K.A., December12.AberrantParalimbicGrayMatterinCriminalPsychopathy.Journalof AbnormalPsychology.Advanceonlinepublication.doi:10.1037/a0026371 [15]M.E.Shenton&H.M.Hamoda&J.S.Schneiderman&S.Bouix&O.Pasternak& Y.Rathi&M.-A.Vu&M.P.Purohit&K.Helmer&I.Koerte&A.P.Lin&C.-F. Westin&R.Kikinis&M.Kubicki&R.A.Stern&R.Zafonte.Areviewof magneticresonanceimaginganddiffusiontensorimagingndingsinmildtraumatic braininjury.BrainImagingandBehavior6:137DOI10.1007/s11682-0129156-5 [16]Restak,Richard.FixingtheBrain.MysteriesoftheMind.Washington,D.C.: NationalGeographicSociety.ISBN0-7922-7941-7. [17]Beaulieu,Christian.Thebasisofanistropicwaterdiffusioninthenervoussystem-atechinicalreview.NMRINBIOMEDICINENMRBiomed.2002;15:435 PublishedonlineinWileyInterSciencewww.interscience.wiley.com. DOI:10.1002/nbm.782 [18]G o tzThomalla,VolkmarGlauche,MartinA.Koch,ChristianBeaulieu DiffusiontensorimagingdetectsearlyWalleriandegenerationofthe pyramidaltractafterischemicstroke.-NeuroImage221767. doi:10.1016/j.neuroimage.2004.03.041 [19]YuchenXie,JeffreyHo,andBabaC.VemuriNonnegativeFactorizationof DiffusionTensorImagesandItsApplications [20]Macovski,A.,NoiseinMRI.MagnResonMed,36:494.doi: 10.1002/mrm.1910360327 [21]Ocali,O.andAtalar,E.,Ultimateintrinsicsignal-to-noiseratioinMRI.Magn ResonMed,39:462.doi:10.1002/mrm.1910390317 [22]Z.Chu,E.A.Wilde,J.V.Hunter,S.R.McCauley,E.D.Bigler,M.Troyanskaya,R. Yallampalli,J.M.Chia,andH.S.Levin,Voxel-BasedAnalysisofDiffusion TensorImaginginMildTraumaticBrainInjuryinAdolescents-AJNRAmJournal ofNeuroradiology31:340-346originallypublishedonlineonDecember3,2009, 10.3174/ajnr.A1806. [23]JonathanM.Silver,M.D.,DiffusionTensorImagingandMildTraumaticBrain InjuryinSoldiers:AbnormalFindings,UncertainImplications-AmJPsychiatry 2012;169:1230-1232.10.1176/appi.ajp.2012.12091230 [24]Niogi,SumitN.PhD;Mukherjee,PratikMD,PhD,DiffusionTensorImaging ofMildTraumaticBrainInjury-JournalofHeadTraumaRehabilitation:Volume 25,July/August2010,p241. [25]Niogi,S.N.P.Mukherjee,J.Ghajar,C.Johnson,R.A.Kolster,R.Sarkar,H.Lee, M.Meeker,R.D.Zimmerman,G.T.Manley,andB.D.McCandliss'Extent ofMicrostructuralWhiteMatterInjuryinPostconcussiveSyndromeCorrelateswith ImpairedCognitiveReactionTime:A3TDiffusionTensorImagingStudyofMild TraumaticBrainInjury-AJNRAmJNeuroradiology29:967-973originallypublishedonlineonFebruary13,2008,10.3174/ajnr.A0970. 75

PAGE 81

[26]LauraMiles,RobertI.Grossman,GlynJohnson,JamesS.Babb,LeonardDiller,and MatildeInglese,Short-termDTIpredictorsofcognitivedysfunctioninmild traumaticbraininjury-BrainInjury200822:2,115-122 [27]JeffreyJ.Bazarian,JianhuiZhong,BrianBlyth,TongZhu,VoykoKavcic,andDerickPeterson.DiffusionTensorImagingDetectsClinicallyImportantAxonal DamageafterMildTraumaticBrainInjury:APilotStudy-JournalofNeurotrauma. September2007,24:1447-1459.doi:10.1089/neu.2007.0241. [28]MarilynF.Kraus,TeresaSusmaras,BenjaminP.Caughlin,CoreyJ.Walker,John A.Sweeney,andDeborahM.Little,Whitematterintegrityandcognitionin chronictraumaticbraininjury:adiffusiontensorimagingstudy-Brain130: 2508-2519rstpublishedonlineSeptember14,2007doi:10.1093/brain/awm216 [29]Ming-XiongHuang,RebeccaJ.Theilmann,AshleyRobb,AnnemarieAngeles, SharonNichols,AngelaDrake,JohnD'Andrea,MichaelLevy,MartinHolland,Tao Song,ShengGe,EricHwang,KevinYoo,LiCui,DewleenG.Baker,DorisTrauner, RaulCoimbra,andRolandR.Lee.IntegratedImagingApproachwithMEG andDTItoDetectMildTraumaticBrainInjuryinMilitaryandCivilianPatientsJournalofNeurotrauma.26:1213-1226.doi:10.1089/neu.2008.0672. [30]SumitN.Niogi,PratikMukherjee,JamshidGhajar,CarlE.Johnson,RachelKolster,HanaLee,MinahSuh,RobertD.Zimmerman,GeoffreyT.Manley,andBruce D.McCandliss,Structuraldissociationofattentionalcontrolandmemoryin adultswithandwithoutmildtraumaticbraininjury-Brain131:3209-3221rst publishedonlineOctober24,2008doi:10.1093/brain/awn247 [31]LangloisJA,Rutland-BrownW,ThomasKE.,TraumaticBrainInjuryinthe UnitedStates:EmergencyDepartmentVisits,Hospitalizations,andDeaths.-Atlanta: CentersforDiseaseControlandPrevention,NationalCenterforInjuryPreventionand Control [32]Langlois,JA.ScD,MPH;Rutland-Brown,WesleyMPH;Wald,MarlenaM. MLS,MPH,TheEpidemiologyandImpactofTraumaticBrainInjury:A BriefOverview-JournalofHeadTraumaRehabilitation.Volume21,September/October2006,p375 [33]D.R.Rutgers,F.Toulgoat,J.Cazejust,P.Fillard,P.Lasjaunias,andD.Ducreux ,WhiteMatterAbnormalitiesinMildTraumaticBrainInjury:ADiffusion TensorImagingStudy-AJNRAmJNeuroradiolMarch200829:514-519originally publishedonlineonNovember26,2007,10.3174/ajnr.A0856. [34]http://www.millenniumhealthcenters.com/sitebuildercontent/sitebuilderpictures/.pond/DTIBreaksAxons.jpg.w560h243.jpg [35]ThomasW.McAllister,MollyB.Sparling,LauraA.Flashman,AndrewJ.Saykin ,NeuroimagingFindingsinMildTraumaticBrainInjury-JournalofClinical andExperimentalNeuropsychologyVol.23,Iss.6,2001 [36]B.Voller,T.Benke,K.Benedetto,P.Schnider,E.Auff,F.Aichner,Neuropsychological,MRIandEEGndingsafterverymildtraumaticbraininjury-Brain Injury13:10,821-827 76

PAGE 82

[37]DavidG.Hughes,AlanJackson,DamonL.Mason,ElizabethBerry,SallyHollis, DavidW.Yates,Abnormalitiesonmagneticresonanceimagingseenacutely followingmildtraumaticbraininjury:correlationwithneuropsychologicaltestsand delayedrecovery-Neuroradiology,Volume46,Issue7,pp550-558 [38]B.A.Cohen,M.Inglese,H.Rusinek,J.S.Babb,R.I.Grossman,andO.Gonen, ProtonMRSpectroscopyandMRI-VolumetryinMildTraumaticBrainInjury-AJNR AmJNeuroradiol28:907-913 [39]HanaLee,MaxWintermark,AlisaD.Gean,JamshidGhajar,GeoffreyT.Manley, andPratikMukherjee,FocalLesionsinAcuteMildTraumaticBrainInjury andNeurocognitiveOutcome:CTversus3TMRIJournalofNeurotrauma.25: 1049-1056.doi:10.1089/neu.2008.0566. [40]HeatherG.Belanger,Ph.D.;RodneyD.Vanderploeg,Ph.D.;GlennCurtiss,Ph.D.; DeborahL.Warden,M.D.RecentNeuroimagingTechniquesinMildTraumaticBrainInjury-TheJournalofNeuropsychiatryandClinicalNeurosciences; 19:5-20. [41]C.E.ShannonAmathematicaltheoryofcommunication,BellSystemTechnicalJournal,vol.27,pp.379-423and623-656. [42]AlbensiBC,KnoblachSM,ChewBG,O'ReillyMP,FadenAI,PekarJJ.DiffusionandhighresolutionMRIoftraumaticbraininjuryinrats:timecourseand correlationwithhistology.ExpNeurol;162:61. [43]ArfanakisK,HaughtonVM,CarewJD,RogersBP,DempseyRJ,MeyerandME. DiffusiontensorMRimagingindiffuseaxonalinjury.AmJNeuroradiol;23: 794. [44]BarkhoudarianG,HovdaDA,GizaCC.Themolecularpathophysiologyof concussivebraininjury.ClinSportsMed;30:33. [45]BazarianJJ,ZhongJ,BlythB,ZhuT,KavcicV,PetersonD.Diffusiontensorimaging detectsclinicallyimportantaxonaldamageaftermildtraumaticbraininjury:apilot study.JNeurotrauma2007;24:1447. [46]BendlinBB,RiesML,LazarM,AlexanderAL,DempseyRJ,RowleyHA,etal. Longitudinalchangesinpatientswithtraumaticbraininjuryassessedwithdiffusiontensorandvolumetricimaging.Neuroimage2008;42:503. [47]BiglerED.NeuroimaginginMildTraumaticBrainInjury.PsycholInjandLaw2010; 2010:36. [48]BlumbergsPC,ScottG,ManavisJ,WainwrightH,SimpsonDA,McLeanAJ.Stainingofamyloidprecursorproteintostudyaxonaldamageinmildheadinjury.Lancet 1994;344:1055. [49]BrowneKD,ChenXH,MeaneyDF,SmithDH.MildtraumaticbraininjuryanddiffuseaxonalinjuryinSwine.JNeurotrauma2011;28:1747. 77

PAGE 83

[50]BuddeMD,JanesL,GoldE,TurtzoLC,FrankJA.Thecontributionofgliosistodiffusiontensoranisotropyandtractographyfollowingtraumaticbraininjury:validation intheratusingFourieranalysisofstainedtissuesections.Brain2011;134:2248. [51]CubonVA,PutukianM,BoyerC,DettwilerA.Adiffusiontensorimagingstudy onthewhitematterskeletoninindividualswithsports-relatedconcussion.JNeurotrauma2011;28:189. [52]DavenportND,LimKO,ArmstrongMT,SponheimSR.Diffuseandspatiallyvariablewhitematterdisruptionsareassociatedwithblast-relatedmildtraumaticbraininjury.Neuroimage;59:2017. [53]FaulM,XuL,WaldMM,CoronadoVG.TraumaticbraininjuryintheUnited States:emergencydepartmentvisits,hospitalizations,anddeaths.CentersforDisease ControlandPrevention [54]HogeCW,McGurkD,ThomasJL,CoxAL,EngelCC,CastroCA.Mildtraumatic braininjuryinU.S.SoldiersreturningfromIraq.NEnglJMed2008;358:453. [55]HughesDG,JacksonA,MasonDL,BerryE,HollisS,YatesDW.Abnormalitieson magneticresonanceimagingseenacutelyfollowingmildtraumaticbraininjury:correlationwithneuropsychologicaltestsanddelayedrecovery.Neuroradiology2004; 46:550. [56]HuismanTA,SchwammLH,SchaeferPW,KoroshetzWJ,Shetty-AlvaN,OzsunarY, etal.Diffusiontensorimagingaspotentialbiomarkerofwhitematterinjuryindiffuse axonalinjury.AJNRAmJNeuroradiol2004;25:370. [57]IngleseM,MakaniS,JohnsonG,CohenBA,SilverJA,GonenO,etal.Diffuseaxonalinjuryinmildtraumaticbraininjury:adiffusiontensorimagingstudy.J Neurosurg;103:298. [58]IversonGL.Complicatedvsuncomplicatedmildtraumaticbraininjury:acute neuropsychologicaloutcome.BrainInj;20:1335. [59]IversonGL.Outcomefrommildtraumaticbraininjury.CurrOpinPsychiatry; 18:301. [60]JacobsB,BeemsT,StulemeijerM,vanVugtAB,vanderVlietTM,BormGF,etal. Outcomepredictioninmildtraumaticbraininjury:ageandclinicalvariables arestrongerpredictorsthanCTabnormalities.JNeurotrauma2010;27:655. [61]KumarR,HusainM,GuptaRK,HasanKM,HarisM,AgarwalAK,etal. Serialchangesinthewhitematterdiffusiontensorimagingmetricsinmoderatetraumaticbraininjuryandcorrelationwithneuro-cognitivefunction.JNeurotrauma2009; 26:481. [62]LangeRT,IversonGL,FranzenMD.Neuropsychologicalfunctioningfollowingcomplicatedvs.uncomplicatedmildtraumaticbraininjury.BrainInj;23:83. [63]LingJ,MeridethF,CaprihanA,PenaA,TeshibaT,MayerAR.Headinjury orheadmotion?Assessmentandquanticationofmotionartifactsindiffusiontensor imagingstudies.HumBrainMapp2012;33:50. 78

PAGE 84

[64]MesseA,CaplainS,ParadotG,GarrigueD,MineoJF,SotoAG,etal.Diffusiontensorimagingandwhitematterlesionsatthesubacutestageinmildtraumatic braininjurywithpersistentneurobehavioralimpairment.HumBrainMapp2011;32: 999. [65]ReevesTM,PhillipsLL,PovlishockJT.Myelinatedandunmyelinatedaxons ofthecorpuscallosumdifferinvulnerabilityandfunctionalrecoveryfollowingtraumaticbraininjury.ExpNeurol;196:126. [66]RosenblumWI.Cytotoxicedema:monitoringitsmagnitudeandcontribution tobrainswelling.JNeuropatholExpNeurol2007;66:771. [67]SmithSM,JenkinsonM,Johansen-BergH,RueckertD,NicholsTE,MackayCE,et al.Tract-basedspatialstatistics:voxelwiseanalysisofmulti-subjectdiffusiondata. Neuroimage2006;31:1487. [68]VanHeckeW,SijbersJ,DeBackerS,PootD,ParizelPM,LeemansA.Ontheconstructionofagroundtruthframeworkforevaluatingvoxel-baseddiffusiontensorMRI analysismethods.Neuroimage2009;46:692. [69]WildeEA,McCauleySR,HunterJV,BiglerED,ChuZ,WangZJ,etal..Diffusiontensorimagingofacutemildtraumaticbraininjuryinadolescents.Neurology; 70:948. [70]JosefM.Ling,AmandaPea,RonaldA.Yeo,FlanneryL.Merideth,StefanKlimaj, CharlesGasparovic,andAndrewR.MayerBiomarkersofincreaseddiffusion anisotropyinsemi-acutemildtraumaticbraininjury:alongitudinalperspectiveBrain135:1281-1292doi:10.1093/brain/aws073 [71]ZDing,JCGore,AWAndersonReductionofNoiseinDiffusionTensorImagesUsingAnisotropicSmoothing,Mag.Res.Med.,53:485-490,2005 [72]JWeickert,HScharrASchemeforCoherence-EnhancingDiffusionFiltering withOptimizedRotationInvariance,CVGPRGroupTechnicalReportattheDepartmentofMathematicsandComputerScience,UniversityofMannheim,Germany. [73]JohnAshburnerandKarlJ.Friston."Voxel-BasedMorphometryTheMethods".NeuroImage11:805. [74]YanagawaY,TsushimaY,TokumaruA,Un-noY,SakamotoT,etal.AquantitativeanalysisofheadinjuryusingT2*-weightedgradient-echoimaging.JTrauma 49:272. [75]MannionRJ,CrossJ,BradleyP,ColesJP,ChateldD,etal.Mechanism-based MRIclassicationoftraumaticbrainsteminjuryanditsrelationshiptooutcome.J Neurotrauma24:128. [76]LopezAD,MathersCD,EzzatiM,JamisonDT,MurrayCJGlobalandregionalburdenofdiseaseandriskfactors,2001:systematicanalysisofpopulation healthdata.Lancet367:1747. [77]LopezAD,MurrayCCTheglobalburdenofdisease,1990.NatMed4: 1241. 79

PAGE 85

[78]ZhuangfeiChen,LiqianCui,MingliLi,LijunJiang,WeiDeng,XiaohongMa,Qiang Wang,ChaohuaHuang,YingchengWang,DavidA.Collier,QiyongGong,TaoLi. Voxelbasedmorphometricanddiffusiontensorimaginganalysisinmale bipolarpatientswithrst-episodemania-ProgressinNeuro-Psychopharmacology andBiologicalPsychiatry,Volume36,Issue2,Pg231-238,ISSN0278-5846, 10.1016/j.pnpbp.2011.11.002. [79]Filler,Aaron.TheHistory,DevelopmentandImpactofComputedImagingin NeurologicalDiagnosisandNeurosurgery:CT,MRI,andDTI.AvailablefromNature Precedings [80]RichardsTL,HeideAC,TsurudaJS,AlvordEC.Vectoranalysisofdiffusion imagesinexperimentalallergicencephalomyelitis.PresentedatSocietyforMagnetic ResonanceinMedicine;Berlin. [81]CalhounVD,LiuJ,AdaliT.AreviewofgroupICAforfMRIdataandICAfor AQ3jointinferenceofimaging,genetic,andERPdata.Neuroimage;45:S163S172. [82]S.D.Gale,L.Baxter,N.Roundy,S.C.JohnsonTraumaticbraininjuryand greymatterconcentration:apreliminaryvoxelbasedmorphometrystudy.JNeurol NeurosurgPsychiatry.2005Jul;76:984-8. [83]Ashburner,J.,andFriston,K.J.MultimodalimagecoregistrationandpartitioningAuniedframework.NeuroImage6:209. [84]Ashburner,J.,andFriston,K.J.Nonlinearspatialnormalizationusingbasis functions.Hum.BrainMapp.7:254. [85]GrossH,KlingA,HenryG,etal.Localcerebralglucosemetabolisminpatientswithlong-termbehavioralandcognitivedecitsfollowingmildtraumaticbrain injury.JNeuropsychiatryClinNeurosci;8:324. [86]NewtonMR,GreenwoodRJ,BrittonKE,etal.AstudycomparingSPECT withCTandMRIafterclosedheadinjury.JNeurolNeurosurgPsychiatry;55:92. [87]MittlRL,GrossmanRI,HiehleJF,etal.PrevalenceofMRevidenceofdiffuse axonalinjuryinpatientswithmildheadinjuryandnormalheadCTndings.AJNR AmJNeuroradiol;15:1583. [88]FrigonC,JardineDS,WeinbergerE,etal.Fractionofinspiredoxygenin relationtocerebrospinaluidhyperintensityonFLAIRMRimagingofthebrain inchildrenandyoungadultsundergoinganesthesia.AJRAmJRoentgenol;179: 791. [89]HaydelMJ,PrestonCA,MillsTJ,LuberS,BlaudeauE,DeBlieuxPMIndicationsforcomputedtomographyinpatientswithminorheadinjury.NEnglJMed. 2000Jul13;343:100-5. [90]ServadeiF,NasiMT,GiulianiG,CremoniniAM,CenniP,ZappiD,TaylorGS CTprognosticfactorsinacutesubduralhaematomas:thevalueofthe'worst'CTscan. BrJNeurosurg.2000Apr;14:110-6. 80

PAGE 86

[91]DemetriadesD,KuncirE,MurrayJ,VelmahosGC,RheeP,ChanLMortality predictionofheadAbbreviatedInjuryScoreandGlasgowComaScale:analysisof 7,764headinjuries.JAmCollSurg.Aug;199:216-22. [92]VanBovenRW,HarringtonGS,HackneyDB,EbelA,GaugerG,BremnerJD, D'EspositoM,DetreJA,HaackeEM,JackCRJr,JagustWJ,LeBihanD,Mathis CA,MuellerS,MukherjeeP,SchuffN,ChenA,WeinerMW.Advancesin neuroimagingoftraumaticbraininjuryandposttraumaticstressdisorder.JRehabil ResDev.2009;46:717-57. [93]AshburnerJNeelinPCollinsDLEvansAandFristonK.Incorporatingprior knowledgeintoimageregistration.NeuroImage6:344-352 [94]TheManagementofConcussion/mTBIWorkingGroup.VA/DODClinical practiceguidelinesformanagementofconcussion/mildtraumaticbraininjury.J.RehabilResDev.2009;46:CP1CP68. [95]KimuraH,MeaneyDF,McGowanJC,GrossmanRI,LenkinskiRE,RossDT,McIntoshTK,GennarelliTA,SmithDHJComputAssistTomogr.Magnetization transferimagingofdiffuseaxonalinjuryfollowingexperimentalbraininjuryinthe pig:characterizationbymagnetizationtransferratiowithhistopathologiccorrelation. Jul-Aug;20:540-6. [96]PirkoI,FrickeST,JohnsonAJ,RodriguezM,MacuraSIReviewofMagnetic resonanceimaging,microscopy,andspectroscopyofthecentralnervoussystemin experimentalanimals.NeuroRx.Apr;2:250-64. [97]CharlesG.GrossAristotleontheBrain.TheNeurosoientist1:245-250 [98]WilliamKeithChambersGuthrie,,AhistoryofGreekphilosophy,Volume1: TheearlierPresocraticsandthePythagoreans,page173.CambridgeUniversityPress [99]HammoudDA,WassermanBAReviewDiffuseaxonalinjuries:pathophysiologyandimaging.NeuroimagingClinNAm.2002May;12:205-16. [100]http://www.mrn.org/collaborate/imaging-equipment/ [101]FinkelsteinE,CorsoP,MillerTandAssociates.TheIncidenceandEconomic BurdenofInjuriesintheUnitedStates.NewYorkNY:OxfordUniversityPress; [102]JonathanPevsnerLeonardodaVinci'scontributionstoneuroscience. TRENDSinNeurosciencesVol.25No.4April2002 [103]LewHL,CifuDX,SigfordB,etalTeamapproachtodiagnosisandmanagementoftraumaticbraininjuryanditscomorbidities.JRehabilResDevelop;44:7 [104]SandersMJandMcKennaK.2001.Mosby'sParamedicTextbook,2ndrevisedEd. Chapter22,"HeadandFacialTrauma."Mosby. [105]ArminSS,ColohanAR,ZhangJHJune2006."Traumaticsubarachnoidhemorrhage:Ourcurrentunderstandinganditsevolutionoverthepasthalfcentury".NeurologicalResearch28:445.doi:10.1179/016164106X115053.PMID16759448. 81

PAGE 87

[106]ChristineL.MacDonald,AnnM.Johnson,DanaCooper,ElliotC.Nelson,NicoleJ. Werner,JoshuaS.Shimony,AbrahamZ.Snyder,MarcusE.Raichle,JohnR.Witherow,RaymondFang,StephenF.Flaherty,DavidL.BrodyDetectionofBlastRelatedTraumaticBrainInjuryinU.S.MilitaryPersonnel.-NEnglJMed.Author manuscript;availableinPMC2011July29.Publishedinnaleditedformas:NEngl JMed.2011June2;364:2091 [107]MichaelJ.Hogan,RogerT.Staff,BrendanP.Bunting,AlisonD.Murray,TrevorS. Ahearn,IanJ.Deary,LawrenceJ.WhalleyCerebellarbrainvolumeaccounts forvarianceincognitiveperformanceinolderadultsCortex,Volume47,Issue4, Pages441http://dx.doi.org/10.1016/j.cortex.2010.01.001 [108]BarkleyJM,MoralesD,HaymanLA,Diaz-MarchanPJ."Staticneuroimaging intheevaluationofTBI".BrainInjuryMedicine:PrinciplesandPractice.Demos MedicalPublishing.pp.140.ISBN1-888799-93-5. [109]Fabian,TC,Patton,Joe,Croce,Martin,Minard,Gayle,Kudsk,Kenneth, Pritchard,F.."BluntCarotidInjury".AnnalsofSurgery223:513. doi:10.1097/00000658-199605000-00007. [110]http://catalog.nucleusinc.com/generateexhibit.php?ID=1870 [111]http://airto.ccn.ucla.edu/BMCweb/HowTo/AC-PCpix/acline.gif [112]AlfredAschoff,PaulKremer,BahramHashemi,StefanKunzeOctober1999. "Thescientichistoryofhydrocephalusanditstreatment".NeurosurgicalReview Springer22:67[67]. [113]K.A.Kulkarni,N.A.Pandya,T.Hulse:Leptomeningealcyst:acomplicationoftrivialheadinjury.TheInternetJournalofPediatricsandNeonatology.2009Volume9 Number2.DOI:10.5580/14cd [114]ShawCM,AlvordEC,andBerry,RG."Swellingofthebrainfollowingischemicinfarctionwitharterialocclusion".ArchivesofNeurology1:161. [115]Gray,Henry.AnatomyoftheHumanBody.Philadelphia:Lea&Febiger,1918; Bartleby.com,2000.www.bartleby.com/107/. [116]https://en.wikipedia.org/wiki/File:Pr%C3%A4zession2.png [117]PlewesDB,KucharczykW.PhysicsofMRI:aprimer.JMagnResonImaging. 2012May;35:1038-54.doi:10.1002/jmri.23642. [118]SquireLF,NovellineRA.Squire'sfundamentalsofradiologythed..HarvardUniversityPress.ISBN0-674-83339-2. [119]Carr,HermanY.July2004."FieldGradientsinEarlyMRI".Physics TodayAmericanInstituteofPhysics57.Bibcode:2004PhT....57g..83C. doi:10.1063/1.1784322. [120]B.Levine,N.Kovacevic,E.I.Nica,G.Cheung,F.Gao,M.L.Schwartz,andS.E. BlackTheTorontotraumaticbraininjurystudy:Injuryseverityandquantied MRIMarch4,200870:771-778 82

PAGE 88

[121]AAPM/RSNAPhysicsTutorialforResidents:FundamentalPhysicsofMRImaging RadiographicsJuly-August200525:41087-1099;doi:10.1148/rg.254055027 [122]Mikheev,A.,Nevsky,G.,Govindan,S.,Grossman,R.andRusinek,H.,Fully automaticsegmentationofthebrainfromT1-weightedMRIusingBridgeBurneralgorithm.J.Magn.Reson.Imaging,27:1235.doi:10.1002/jmri.21372 83


ERROR LOADING HTML FROM SOURCE (http://ncf.sobek.ufl.edu//design/skins/UFDC/html/footer_item.html)