ContentslistsavailableatScienceDirect
European
Journal
of
Radiology
jo u r n al ho me p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e j r a d
Diffusion
volume
(DV)
measurement
in
endometrial
and
cervical
cancer:
A
new
MRI
parameter
in
the
evaluation
of
the
tumor
grading
and
the
risk
classification
Pier
Paolo
Mainenti
a,∗,
Laura
Micol
Pizzuti
a,
Sabrina
Segreto
b,
Marco
Comerci
a,
Simona
De
Fronzo
b,
Federica
Romano
b,
Vincenzina
Crisci
b,
Michele
Smaldone
b,
Ettore
Laccetti
b,
Giovanni
Storto
c,
Bruno
Alfano
a,
Simone
Maurea
b,
Marco
Salvatore
d,
Leonardo
Pace
eaIBBCNR,Napoli,Italy
bDipartimentodiScienzeBiomedicheAvanzate,SezionediRadiologia,UniversitàdiNapoli“FedericoII”,Napoli,Italy cIRCCSCROB,RioneroinVulture,Italy
dIRCCSSDN,Napoli,Italy
eDipartimentodiMedicinaeChirurgia,UniversitàdegliStudidiSalerno,Salerno,Italy
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received9March2015
Receivedinrevisedform7September2015 Accepted25October2015 Keyword: MRI DWI ADCmaps Diffusionvolume Cervical Endometrial Cancer
a
b
s
t
r
a
c
t
Purpose:AnewMRIparameterrepresentativeofactivetumorburdenisproposed:diffusionvolume (DV),definedasthesumofallthevoxelswithinatumorwithapparentdiffusioncoefficient(ADC)values withinaspecificrange.Theaimsofthestudywere:(a)tocalculateDVonADCmapsinpatientswith cervical/endometrialcancer;(b)tocorrelateDVwithhistologicalgrade(G)andriskclassification;(c)to evaluateintra/inter-observeragreementofDVcalculation.
Materialsandmethods:Fifty-threepatientswithendometrial(n=28)andcervical(n=25)cancers under-wentpelvicMRIwithDWIsequences.Bothendometrialandcervicaltumorswereclassifiedonthebasis ofG(G1/G2/G3)andFIGOstaging(low/medium/high-risk).
Asemi-automatedsegmentationprocedurewasusedtocalculatetheDV.AfreehandclosedROI out-linedthewholevisibletumoronthemostrepresentativesliceofADCmapsdefinedastheslicewiththe maximumdiameterofthesolidneoplasticcomponent.Successively,twothresholdsweregeneratedon thebasisofthemeanandstandarddeviation(SD)oftheADCvalues:lowerthreshold(LT=“meanminus threeSD”)andhigherthreshold(HT=“meanplusoneSD”).TheclosedROIwasexpandedin3D,including allthecontiguousvoxelswithADCvaluesintherangeLT-HT×10–3mm2/s.
AKruskal–WallistestwasusedtoassessthedifferencesinDVamongGandriskgroups.
Intra-/inter-observervariabilityforDVmeasurementwasanalyzedaccordingtothemethodofBland andAltmanandtheintraclass-correlation–coefficient(ICC).
Results:DVvaluesweresignificantlydifferentamongGandriskgroupsinbothendometrial(p<0.05)and cervicalcancers(p≤0.01).Forendometrialcancer,DVofG1(mean±sd:2.81±3.21cc)neoplasmswere significantlylowerthanG2(9.44±9.58cc)andG3(11.96±8.0cc)ones;moreover,DVoflowriskcancers (5.23±8.0cc)weresignificantlylowerthanmedium(7.28±4.3cc)andhighrisk(14.7±9.9cc)ones. Forcervicalcancer,DVofG1(0.31±0.13cc)neoplasmswassignificantlylowerthanG3(40.68±45.65 cc)ones;moreover,DVoflowriskneoplasms(6.98±8.08cc)wassignificantlylowerthanmedium (21.7±17.13cc)andhighrisk(62.9±51.12cc)onesandDVofmediumriskneoplasmswassignificantly lowerthanhighriskones.
Theintra-/inter-observervariabilityforDVmeasurementshowedanexcellentcorrelationforboth cancers(ICC≥0.86).
Conclusions:DVisanaccurateindexfortheassessmentofGandriskclassificationofcervical/endometrial cancerswithlowintra-/inter-observervariability.
©2015ElsevierIrelandLtd.Allrightsreserved.
∗ Correspondingauthorat:CorsoVittorioEmanuele,67080122Napoli,Italy.Fax:+390817616013. E-mailaddress:pierpamainenti@hotmail.com(P.P.Mainenti).
http://dx.doi.org/10.1016/j.ejrad.2015.10.014
1. Introduction
Thetreatmentandprognosisofcarcinomaoftheuterinecervix
and endometrium are determined by the histological subtype,
grade(G)andFIGOstaging.MRIplaysacriticalroleinthe
preop-erativestagingofthesetumorsprovidingdiagnosticinformation
regardingtumorsize,depthofinvasion,infiltrationofcontiguous
organsandlymphnodeinvolvement;evaluationofthese
parame-terscontributeschieflyfortreatmentplanning[1–3].
DiffusionweightedMRimaging(DWI)isafunctionaltechnique
thatlooksattheBrownianmotionofwaterintissues.In
biologi-caltissues,theBrownianmotionisrestrictedbyinteractionswith
cellmembranes and macromoleculesona microscopiclevel as
wellasit is modified byanyarchitectural tissue changes [4,5].
IncreasedtissuecellularityobservedintumorsrestrictsBrownian
motion,whichcanbequantifiedbycalculationoftheapparent
dif-fusioncoefficient(ADC).Inparticular,ADChasbeenshowntobe
inverselycorrelatedwithtumorcellularityandithasbeen
clini-callyappliedtodistinguishbenignfrommalignanttumors,toassess
tumorgrade,todelineatetumorextent,aswellastoevaluatetumor
treatmentresponse[5–17].
PreviousstudiesregardingtheapplicationsofDWIin
gyneco-logicdiseasesdemonstratedthatcervicalcancershavesignificantly
lowerADCvaluescomparedtonormalcervicaltissue[6–9].
Simi-larfindingshavebeenfoundinendometrialcancers,whichshow
lowerADC values than normal endometriumor benign lesions
[10–17].Ontheotherhand,thereisnoconsensusaboutthe
corre-lationofADCvalueswiththeGofthecervicalortheendometrial
tumors: some authors did not find any significant correlation
[11,12,14,18],whileothersfoundsignificantresultsonlyin
differ-entiatingG1fromG3lesionsinendometrialcancers[13].Moreover,
nosignificantdifferencewasobservedamongADCvaluesof
cer-vicalcancerwhentumorsweresubgroupedonthebasisofFIGO
staging[19].Therefore,thepotentialroleofADCvaluesinthe
eval-uationofGandFIGOstagingofthecervicalandendometrialcancer
isstillunclear.
Significantdifferences arereportedin thetechniqueusedto
acquireandtoanalyzeDWI:heterogeneityofacquisition
param-eters (different b-values and acquisition planes), different ROI
placementmethods(singleROI,multipleROIs,manualROIs)and
different ADC values analysis (minimum ADC=ADCmin, mean
ADC=ADCmean,ADCmin/ADCmean=ADCratio)[6,7,10–15].Thus,
lackofacurrentstandardizedDWIprotocolmayhavecontributed
tothelargevarietyofresultsintherelationshipsbetweenADC
valuesandGandFIGOstaging[11–15,18,19].
Methods of ADC maps analysis different from conventional
ROIbasedmeasurementsand abletorepresentthewholesolid
viablecomponentoftheneoplasmmightovercometheabove
mis-matchingresults.Anewstrategy toanalyzetheADCmapsmay
besuggested by PET/CT: a positive correlationbetween tumor
FDGuptakeandthedegreeofcellularityhasbeendemonstrated
[20–22];asaconsequenceofthispositivecorrelation,themetabolic
tumorvolume(MTV),definedasthevolumeoftumortissueswith
FDGuptakehigherthanadeterminatethreshold,hasbeenshown
toestimatetheactivetumorburden[23].Inthesameway,a
param-etersimilartoMTVmaybecomputedonADCmaps.Becauseofthe
inversecorrelationbetweenADCandtumorcellularity,thevolume
oftissueshowingADCvaluesbelowadeterminatethresholdmay
berepresentativeoftheactivetumorburdenofaneoplasm.Thus,
adiffusionvolume(DV)maybedefinedasthesumofallthevoxels
withinatumorwithADCvaluesbelowadeterminatethresholdand
mayrepresenttheactivetumorburden.Sofar,nostudiesdealing
withDVincancerhavebeenpublished.
Therefore,thepurposesofthepresentstudywere(1)todevelop
amethodtocalculatetheDVonADCmapsinpatientswithcervical
and endometrial cancers; (2) to evaluate the intra- and
inter-observeragreementoftheDVcalculation;(3)tocorrelatetheDV
aswellasthecommonlyusedADCvalueswiththehistological
tumorgradeandtheriskclassificationandtocomparedirectlythe
performanceoftheseparameters.
2. Materialsandmethods
2.1. Populationselection
Thisretrospectivestudyenrolledallpatientswithendometrial
orcervicalcancerwhounderwentpelvicMRIwithDWIsequences
atourinstitutionbetweenNovember2010andSeptember2012.
The inclusion criterion was histologically proven cervical or
endometrialcancer.Theexclusioncriterionwasprevious
preop-erativechemo-and/orradio-therapybeforeMRIexamination.
Thestudywasapprovedbylocalresearchethicscommittee.
BeforeMRIscan,allpatientsgavetheirinformedconsentto
per-formexaminationaswellasthepermissionfortheuseoftheir
anonymiseddataforresearchpurposes.Thisprocedurerepresents
astandardprotocolinourinstitution.
Theclinicalpresentation,thehistologicaltype,theGandthe
stagingforeachpatientwereobtainedfromtheclinicalrecordsof
thehospital.Thestagingwasdeterminedaccordingtothe
guide-linesoftheInternationalFederationofGynecologyandObstetrics
(FIGO).
Accordingto Gand FIGO staging, endometrial cancerswere
stratified in three classes of risk determining the management
strategyandprognosis(Table1):lowrisk(stageIAwithG1orG2),
mediumrisk (stageIA withG3or stageIBwithanyGorstage
II)and highrisk (stageIIIor IV)[24,25].Similarly,accordingto
FIGOstaging,cervicalcancerswerestratifiedinthree classesof
riskdeterminingthemanagementstrategyandprognosis(Table1):
lowrisk(earlystagedisease:stageIAorIB1orIIA1),mediumrisk
(earlystagebulkydisease:stageIB2orIIA2)andhighrisk(locally
advanceddisease:stageIIBorIIIorIV)[25,26].Thedefinition“bulky
disease”identifiesearlystage(IBandIIA)neoplasmswithgreatest
diameter>4cm.Anincreasedriskofnodalinvolvementis
associ-atedtotheselesions[25,26].
2.2. MRItechnique
Images were acquired using a 3T MR scanner (Magnetom
TrioSiemens,Erlangen,Germany)equippedwithamultichannel
phasedarraycoil.Forallsequences,thefieldofview(FOV)and
thenumberofsliceswereoptimizedforeach individualpatient
tocovertheanatomyofinterest.Axialandcoronalplaneswere
acquiredperpendicularandparalleltothelongaxisoftheuterine
bodyforendometrialcancerandofcervixforcervicalcancer.
All patients underwent the following MRI unenhanced and
contrast-enhancedprotocol:
-T1-weightedturbospin-echo(TSE)intheaxial-obliqueplane(TR
828ms;TE10ms;slicethickness4mm;matrixsize256×256;
averages3;flipangle140◦;acquisitiontime2.26min);
-T2-weightedTSEinthesagittalplane(TR4500ms;TE102ms;
slicethickness4mm;matrixsize256×256;averages3;
acquisi-tiontime2.26min);
-T2-weightedTSE in theaxial-oblique planewithandwithout
fat suppression(TR 5000ms; TE 94ms; slice thickness 4mm;
matrixsize 256×256;flipangle140◦,averages2;acquisition
time2min);
-T2-weightedTSEinthecoronal-obliqueplane(TR6000ms;TE
106ms;slicethickness4mm;matrixsize256×256;averages2;
Table1
Treatmentandprognosisofendometrialandcervicalcancerstratifiedforclassofrisk.
Endometrialadenocarcinoma Lowrisk Mediumrisk Highrisk
Treatment Totalhysterectomywithbilateral
salpingo-oophorectomy
Radicalhysterectomywithbilateral
salpingo-oophorectomyandpelvic
lymphadenectomy
Para-aorticlymphadenectomymaybe
considered RTisindicated
CTmaybeconsidered
Maximalsurgicaldebulkinginpatientswithagood
performancestatusassociatedtoCT/RT(stageIII) Anteriorandposteriorpelvicexenterationassociated toCT/RT(stageIVa)
PalliativesurgeryassociatedtoCT/RT(stageIVb)
Prognosis Fiveyearssurvivalrate:
>95%
Fiveyearssurvivalrate: 80–90%
Fiveyearssurvivalrate: StageIII:55–65%
StageIV:25–35%
Cervicalcarcinoma Lowrisk Mediumrisk Highrisk
Treatment StageIA
• Conizationortrachelectomy(if
fertilityistobepreserved) • Totalorradicalhysterectomy(if
fertilityisnottobepreserved)
• PLNDifLVSI
StageIB1andIIA1
• RadicalhysterectomyandPLND
• RTmaybeconsidered
RadicalhysterectomyandPLND
associatedtoRTandCT
CombinationofRT/CT
Prognosis Fiveyearssurvivalrate:
StageIA:>90%
StageIB1andIIA1:70–85%
Fiveyearssurvivalrate: 65–75%
Fiveyearssurvivalrate: StageIIB:55–65% StageIII:25–40% StageIV:<15%
-T1-weighted volume interpolated breath hold examination
(VIBE) (TR 3.30ms; TE 1.17ms; slice thickness 2mm; matrix
size256×256;averages1;flipangle13◦;acquisitiontime20s
for each phase) before and after bolus injection of
gadopen-tetate dimeglumine (Magnevist; Berlex Laboratories; dose:
0.2mmol/kg; injection bya powerinjector; rateof injection:
3ml/secfollowedby20mlofnormalsalineflush)at30,60,120s
inthesagittalplaneandat180sintheaxial-obliqueplane;
BeforecontrastenhancedT1-VIBEsequences,DWIwasobtained
using a single-shot-echo-planar imaging sequence under free
breathingwithchemicalshiftselectivefat-suppressiontechnique
(SPAIR)andparallelimagingtechnique(GRAPPA-2).Thefollowing
parameterswereusedintheaxial-obliqueplanes(TR5700ms,TE
69ms,slicethickness4mm;matrixsize128×128;averages5;
b-value0,500and1000s/mm2,acquisitiontime3.07min)andinthe
sagittalplanes(TR5000ms,TE70ms,slicethickness4mm;matrix
size128×128;averages5;b-value0,500and1000s/mm2,
acqui-sition time 3.10min). Followingtheacquisition ofb=0 images,
motion-probinggradientsinthreeorthogonalorientationswere
sequentiallyappliedforeachbvalue.ADCmapwascomputedfrom
theb=0imagesandthediffusion-weightedimagesreconstructed
forb=500and1000s/mm2.Thestandardsoftwareontheimaging
console(SyngoVE36A,Siemens,Erlangen,Germany)wasusedfor
ADCmapcalculation.
2.3. ADCmapsanalysis
Five differentradiologists(eachone withatleast 4years of
experienceinfemalepelvisMRimaging)wereaskedtoanalyse
theimages:threeofthemappliedadifferentROImethodto
calcu-latetheconventionalADCvalues(S.S.,L.P.,M.S.)andtwoofthem
(E.L.,S.D.)definedtheDV.Thechoicetoselectfiveradiologistswas
dictatedbythenecessitytopreventthattheapplicationofeach
methodofanalysismightbeinfluencedbytheothers.Inorderto
performinter-andintra-observeranalysisofDV,tworadiologists
wereinvolvedandoneofthemwasaskedtorepeattheprocedure
onemonthlater(S.D.).
AllMRIsequencesperformedwereavailabletothereadersatthe
timeofADCmapsanalysis.Eachreaderwasaskedtoidentifythe
tumor,todefineitscontoursandtorecognizethecystic/necrotic
portionsofthelesion.Eachreaderusedtheunenhanced,
contrast-enhancedandDWIimagestodefinethesolidpartofthetumor
anditscontours.Thecystic/necroticportionsweredefinedasareas
withfluid-likehyperintensesignalonT2imagesandDWIimages
obtainedatb0valueandwithoutenhancementonT1VIBElesions.
Eachsequencecouldbeinspectedindividuallyaswellas
compara-tivelywiththeothers;moreover,MRIfused-imageswereobtained
ifnecessary. TheROIs placementwas madeusingvisual
corre-lationof theADCmapswithinformation onthecorresponding
unenhanced,contrast-enhancedandDWIimages.Thereaderswere
blindedtoeachother’sresults,clinicalpatientdataandpathology
reports.
2.3.1. ADCvalues
ThreedifferentradiologistsanalyzedtheADCmapsapplying
eachoneamethod.ThemethodA(S.S.)consistedinplacinga
sin-gularROI,aslargeaspossible,withinthetumoronADCmaps,in
theslicecontainingthelargestsolidtumorarea.ThemethodB(L.P.)
consistedindrawingfivecircularROIs,wherepossible,
approxi-matelyof5mm2,withinthetumoronADCmaps,ineachslicein
whichthetumorwasdetectable.ThemethodC(M.S.)consistedin
outliningoneROIwhichincludedthewholeneoplasticareaonADC
maps,ineachsliceinwhichthetumorwasdetectable.
For eachpatienttheADCvalues werecalculatedbothin the
axial-obliqueandthesagittalplanes.
TheROIswerealwaysplacedinthesolidcomponentsofthe
tumors,avoidingcystic/necroticareas.
TheADCvalueswereobtainedintermsofADC×10–3mm2/s
(millimeterssquared/second).
ADCmin, ADCmean and ADCratio were calculated for each
method.
When using the method A, ADCmin and ADCmean were
extractedinthesingularROI.ForthemethodB,ADCminwasthe
minimumvalueofADCobtainedamongallthecircularplacedROIs,
Fig.1.a,b,candd.Calculationofdiffusionvolume(DV)inapatientwithendometrialcancer.ThemostrepresentativesliceoftheneoplasmwasselectedonADCmaps(a). Thesolidlesionwasoutlinedmanuallywithaline;meanandstandarddeviation(SD)oftheADCvalueswerecalculatedontheresultingfreehandROI;twothresholdswere
successivelygenerated:lowerthreshold(LT)definedas“meanvalueminusthreeSD”andhigherthreshold(HT)definedas“meanvalueplusoneSD”(b).Theclosedsurface
wasdefinedonthebasisofLTandHT(c).Successively,theclosedsurfacewasexpandedin3DincludingallthecontiguousvoxelswithADCvaluesintherangeLT–HT,as
aresulttheDVexpressedinccwasobtained(d).Infigured,thedarkpartofthevolumerepresentstheleftsideofthetumorrespecttotherepresentativeslice,whilethe greenpartindicatestherightside.
extractedbyeachROI.InthemethodC,ADCminwasconsidered
astheminimumvalueofthetotalsliceswithinthetumor,while
ADCmeanwascalculatedastheaverageofmeanADCvaluesineach
slice.
2.3.2. Diffusionvolume(DV)
Asemi-automatedsegmentationprocedurewritteninMatlab
(TheMatworks)wasusedforimageanalysis.Thegoalofthe
seg-mentationprocedurewastoselectthesolidcellularpartsofthe
neoplasmandtoexcludethecystic/necrotic/inflammatoryones.
Theradiologistselectedthemostrepresentativeslice,definedas
theslicewiththemaximumdiameterofthesolidcomponentofthe
tumor,andsuccessivelyoutlinedmanuallythewholevisiblesolid
neoplasticcomponentwitha line.Meanandstandarddeviation
(SD)oftheADCvalueswerecalculatedontheresultingfreehand
ROI.Twothresholdsweresuccessivelygenerated:lowerthreshold
(LT)definedas“meanvalueminusthreeSD”andhigher
thresh-old(HT)definedas“meanvalueplusoneSD”.Theclosedsurface
representedbythefreehandROIwasexpandedin3D,including,
iteratively,allthecontiguousvoxelswithADCvaluesintherange
LT-HT.Attheendofprocess,theoperatorcheckedthefinal3DROI
toverifythatnovoxeloutsidethecontoursofthesolid
neoplas-ticcomponentwasincluded,eventuallydeletedthemisclassified
voxelsandeditedthefinal3D-ROI(Fig.1).
Asstatedabove,thecystic/necroticportionswereeliminatedon
thebasisofthevisualcomparisonofADCmapswithT2-weighted
andcontrast-enhancedimages,howeverweestimatedthatamore
restrictivequantitativelyapproachontheADCmapspreventedthe
erroneousinclusionofcomponentsdifferentfromsolidportions
ofthetumor.Theasymmetricrangeswitchedtowardthelower
ADCvalueswassuggestedbytheinversecorrelationbetweenADC
and tumorcellularity(the lower ADCcorrespondtothehigher
cellulardensityvoxels)aswellasthelowervaluesofADCfor
malig-nantcervicalandendometrialtumorrespecttonormaltissuesor
benignlesions.Asaresult,theHTrepresentedafurtherstrategy
toexcludeinflammatorychangesaccompanyingthetumoraswell
ascystic/necroticcomponentsnotidentifiedbyvisualcomparison
withunenhancedandcontrast-enhancedMRimagesandto
pre-venttheerroneousinclusionofcontiguousnormaltissues.TheLT
wasintendedtoexcludethenoisyvoxels.
Thewholeprocedurewasperformedforboththesagittaland
theaxialimagesbytworadiologistsforinter-observervariability
(E.L.andS.D.).Oneofthetwoobserversrepeatedtheprocedure
onemonthlatertoevaluateintra-observervariabilityoftheresults
(S.D.).
Onlytheprimarytumorswereincludedintheanalysis,andthus
pelviclymphnodeswerenotconsideredforDVanalysis.
2.4. MorphologicalvolumecalculatedonT2images
Anotherradiologist(F.R.)withatleast4yearsofexperiencein
femalepelvisMRimagingoutlinedthewholevisibletumorwith
aclosedlineineachsliceinwhichthetumorwasdetectable.The
closedlineincludedonlythesolidportionsofthetumors,avoiding
fluid/necrotic portions.The T2 volume wasautomatically
com-putedonthebasisofeachslicearea,theslicethicknessandthe
inter-slicegap.
The procedurewas executed for both the sagittal and axial
Fig.2.a,b,candd.MeanandstandarddeviationofDVvalues(expressedincc)ofendometrialcancersontheyaxis(aandc:axialDV;bandd:sagittalDV)accordingto grading(aandb)andriskgroupclassification(candd)reportedonthexaxis.
TheT2volumewasconsideredequaltozerointhosepatientsin
whomthelesionwasnotdefinedonT2imagesanditsidentification
couldbeobtainedthankstotheothersetsofimages.
2.5. Statisticalanalysis
Kruskal–Wallistestwithpost-hocanalysiswasusedtoassess
thestatisticaldifferences inmeasuredparameters among
grad-inggroups as wellas among risk groups a nonparametric.A p
value<0.05wasconsideredsignificant.
Intra-and inter-observervariabilityfor tumorvolume
mea-surementswasanalyzed accordingtothemethodofBlandand
Altmanandbycalculatingintraclasscorrelationcoefficient(ICC)
(0.00–0.2poor,0.21–04fair,0.41–0.6moderate,0.61–0.80good
and0.81–1.00excellentcorrelation)withcorresponding95%
con-fidenceinterval(CI).Inthesameway,thedifferencebetweenaxial
andsagittalmeasurementswasevaluated.
Receiveroperatorcharacteristic(ROC)curveanalysiswas
per-formedtoevaluatetheimpactofthedifferentparametersonboth
Gandriskclassification.ROCcurveswerefittedforeach
param-eterandtask, and theareaunderthecurve(AUC)wasusedas
performanceindex.The95%confidenceintervalfortheareawas
usedtotestthehypothesisthatthetheoreticalareawas0.5.Ifthe
confidenceintervaldidnotincludethe0.5value,thentherewas
evidencethatthetesthadanabilitytodistinguishbetweenthetwo
groups.Apvalue<0.05wasconsideredsignificant.Themethodof
Delongwasusedinordertocomparingtheareasundercorrelated
receiveroperatingcharacteristiccurves.
MedCalcStatisticalSoftwareversion13.1.2wasusedfor
sta-tisticalanalysis(MedCalcSoftwarebvba,Ostend,Belgium;http://
www.medcalc.org;2014).
3. Results
3.1. Patients
Thestudy groupconsistedof53 patients, 28 with
histologi-callyprovenendometrialcancer(agerange,31–85years;meanage
58±117years)and25withhistologicallyprovencervicalcancer
(agerange,22–73years;meanage47±121years).
Thirty-twopatientswerepostmenopausal(21with
endome-trialcancer,11withcervicalcancer).
TwoclinicalscenarioswereobservedbeforeMRI:(1)patients
with highly suspicious lesions at transvaginal ultrasonography
examination (n28);(2) patientswithpositive hysteroscopyfor
Table2
Clinicalpresentation,histologicaltype,grade,FIGOstagingandclassesofriskofendometrialandcervicalcarcinoma.
Endometrialadenocarcinoma n◦ Cervicalcarcinoma n◦
Clinicalpresentation Metrorrhagia 9 Metrorrhagia 9
Menometrorrhagia 2 Menometrorrhagia 2
Vaginalbleeding 12 Vaginalbleeding 7
Asymptomatic 5 Asymptomatic 7
Histologicaltype Endometrioid 26 Squamous 25
Sarcomatoid 1
Squamous 1
Grade Grade1 8 Grade1 2
Grade2 13 Grade2 3
Grade3* 7 Grade3 20
FIGOstaging IA 14 IA2 2
IB 11 IB1 7 IIIA 1 IB2 1 IIIC1 2 IIA1 2 IIA2 2 IIB 5 IIIB 4 IVA 1 IVB 1
Classesofrisk Lowrisk 13 Lowrisk 11
Mediumrisk 8 Mediumrisk 3
Highrisk 7 Highrisk 11
* Thesarcomatoidandsquamousadenocarcinomasweretakentorepresenttheequivalentofgrade3endometrioidadenocarcinomasbecausetheyareconsideredtobe
aggressivehistologicalsubtypesandareassociatedwithaworseprognosisthanendometrioidadenocarcinoma.
Table3
Mean,standarddeviation(sd)andrangeofDVandT2volumeforendometrialandcervicalcancer.DVandT2volumeareexpressedincubiccentimeter.
Endometrialcancer Cervicalcancer
Axial Sagittal Axial Sagittal
Diffusionvolume(DV) Mean±sd 9.74±8.99 10.26±9.21 33.33±36.37 32.82±34.23 Range 0.45–33.66 0.39–35.81 0.20–205.33 0.19–181.67 T2volume Mean±sd 12.4±11.5 12.6±11.5 36.86±35.6 35.69±34.82 Range 0–36.45 0–36.8 0–215 0–198
No therapy (surgery, chemo-radiotherapy) was performed
beforeMRIstudy.
Theclinicalpresentation,thehistologicaltype,theG,theFIGO
stagingandtheriskclassificationclassesaresummarizedinTable2.
3.2. Images
Theprimarytumorwasdefinedatleastinonesetofimagesinall
28patientswithendometrialcancerand25patientswithcervical
cancers.
3.2.1. Endometrialcancer
MeanvaluesandrangevaluesofDV,ADCmin,ADCmeanand
ADCratiofor eachROI protocoland T2volumearedisplayed in
Tables3and4.
BothaxialandsagittalDVvaluesweresignificantly(p=0.012;
p=0,021respectively)differentamongGgradegroups;in
particu-lar,post-hocanalysisrevealedasignificantdifferenceofG1vsG2
andG3(Fig.2aandFig.2b).Similarly,bothaxialandsagittalDV
valuesweresignificantly(p=0.014;p=0.02respectively)different
amongriskgroups;inparticular,post-hocanalysisrevealeda
sig-nificantdifferenceofthelowriskversusthemediumandthehigh
risk(Fig.2candFig.2d).Moreover,theDVmeasuresobtainedon
axialandsagittalplanesshowedanexcellentcorrelation(Fig.3).
Noneoftheothervariables(ADCmin,ADCmean,ADCratio)using
anyof themethodA, Band C showedsignificant differenceat
Kruskal–Wallisanalysis.
Both axial and sagittal T2 volumes were not significant at
Kruskal–WallisanalysisamongG-gradegroupsandriskgroups.
Fig. 3. Inter-planes reproducibility for diffusion volume (DV) measurements (expressedincc)performedinaxialandsagittalplanesfortheendometrial can-cer.Bland–AltmanplotsoftheaverageofaxialandsagittalDV(xaxis)againstthe differencebetweenaxialDVandsagittalDV(yaxis).Thecontinuouslinerepresents themeanabsolutedifferenceinDVbetweenthetwoplanes;thedashedlines rep-resentthe95%confidenceintervalsofthemeandifferences.TheICCwasexcellent (ICC:0.94;95%CI:0.93–0.99).
3.2.2. Cervicalcancer
Mean values, standard deviation and range of DV, ADCmin,
ADCmeanandADCratioforeachROIprotocolandT2volumeare
Fig.4.a,b,candd.MeanandstandarddeviationofDVvalues(expressedincc)ofcervicalcancersontheyaxis(aandc:axialDV;bandd:sagittalDV)accordingtograding (aandb)andriskgroupclassification(candd)reportedonthexaxis.
Fig.5.Inter-planesreproducibilityfordiffusionvolume(DV)measurements per-formedinaxialandsagittalplanesforcervicalcancer.Bland–Altmanplotsofthe averageofaxialandsagittalDV(xaxis)againstthedifferencebetweenaxialDVand sagittalDV(yaxis).Thecontinuouslinerepresentsthemeanabsolutedifferencein DVbetweenthetwoplanes;thedashedlinesrepresentthe95%confidenceintervals ofthemeandifferences.TheICCwasexcellent(ICC:0.97;95%CI:0.95–0.99).
BothaxialandsagittalDVvalues weresignificantly(p=0.01)
differentamongGgradegroups;inparticular,post-hocanalysis
revealedasignificantdifferencebetweenG1andG3(Fig.4aand
b).Similarly,bothaxialandsagittalDVvaluesweresignificantly
(p<0.001)differentamongalltheriskgroups(Fig.4candd).
More-over,theDVmeasuresobtainedonbothaxialandsagittalplanes
showedanexcellentcorrelation(Fig.5).
Relatively to the method A, the axial ADCmean
(G1: 1.17×10–3mm2/s; G2: 0.84×10–3mm2/s; G3:
0.75×10–3mm2/s)andtheaxialADCmin(G10.96×10–3mm2/s;
G2:0.75×10–3mm2/s;G3:0.56×10–3mm2/s)weresignificantly
(p<0.05)differentamongGgradegroups;inparticular,post-hoc
analysisrevealedasignificantdifferenceofG1vsG3.Noneofthe
variables ofmethodA wasable todiscriminateamongtherisk
groups.
Relativelyto themethod B, theaxial ADCmean (earlystage
1.04×10–3mm2/s;earlystagebulkydisease:0.90×10–3mm2/s;
locallyadvanceddisease:0.79×10–3mm2/s)wassignificantly
dif-ferent (p=0.01) among the risk groups; in particular, post-hoc
analysisrevealedasignificantdifferenceoftheearlystagevsthe
earlystagebulkydiseaseandthelocallyadvanceddisease.None
ofthevariablesofmethodBwasabletodiscriminateamongtheG
Table4
Mean,standarddeviation(sd)andrangeofADCmean,ADCmin,ADCratioofeachmethodforendometrialandcervicalcancer.ADCminandADCmeanareexpressedin 10−3mm2/s.
Endometrialcancer Cervicalcancer
Axial Sagittal Axial Sagittal
MethodA ADCmean Mean±sd 0.86±0.17 0.87±0.16 0.88±0.16 0.89±0.20 Range 0.58–1.30 0.55–1.20 0.57–1.37 0.56–1.53 ADCmin Mean±sd 0.68±0.16 0.67±0.13 0.73±0.15 0.47±0.17 Range 0.37–1.10 0.42–0.91 0.37–1.08 0.44–1.25 ADCratio Mean±sd 0.79±0.06 0.77±0.07 0.83±0.06 0.83±0.04 Range 0.56–0.89 0.6–0,99 0.66–0.98 0.77–0,90 MethodB ADCmean Mean±sd 0.93±0.16 0.95±0.17 0.94±0.15 0.97±0.20 Range 0.70–1.3 0.74–1.46 0.69–1.33 0.66–1.33 ADCmin Mean±sd 0.63±0.15 0.61±0.16 0.70±0.16 0.70±0.17 Range 0.37–1.09 0.42–1.21 0.39–1.06 0.44–1.00 ADCratio Mean±sd 0.68±0.10 0.65±0.10 0.73±0.07 0.72±0.08 Range 0.53–0.84 0.57–0.79 0.56–0.80 0.66–0.76 MethodC ADCmean Mean±sd 0.95±0.18 0.96±0.18 0.98±0.11 1.00±0.12 Range 0.68–1.52 0.68–1.49 0.73–1.14 0.72–1.10 ADCmin Mean±sd 0.66±0.20 0.66±0.19 0.66±0.13 0.67±0.13 Range 0.37–1.26 0.40–1.27 0.39–0.86 0.44–0.88 ADCratio Mean±sd 0.68±0.11 0.68±0.10 0.67±0.08 0.67±0.07 Range 0.42–0.83 0.42–0.86 0.52–0.84 0.53–0.79
Relativelytothe C method, noneof the variables (ADCmin,
ADCmean,ADCratio)wassignificantdifferentatKruskal–Wallis
analysis.
Kruskal–Wallisanalysisdidnotshowanysignificantdifference
amongGgradeforbothaxialandsagittalT2volumes.Onthe
con-trary,bothaxialandsagittalT2volumesweresignificantly(p=0.01)
differentamongtheriskgroups;inparticular,post-hocanalysis
differentiatedsignificantlythelocallyadvanceddiseasefromthe
earlystageandtheearlystagebulkydiseaseonaxialimagesand
thelocallyadvanceddiseasefromtheearlystageonsagittalimages.
3.2.3. Intra-andinter-observervariabilityofdiffusionvolume.
Theintra-observervariabilityandtheinter-observervariability
wereexcellent(ICC>0.81)relativelytobothendometrial(Fig.6)
andcervicalcancer(Fig.7).
3.2.4. ROCanalysis
Inordertoevaluateandcomparetheclinicalimpact,ifany,of
thedifferentparametersevaluated,ROCanalysiswasperformed
forbothGandriskclassification.Actually,patientswithG1andG2
cancersweregrouped,aswellas,patientswithlowandmedium
classofrisk.
Inendometrialcancergroup,theROC-AUCforDVwas
statisti-callysignificantforbothGandriskclassification(Tables5and6),
onthecontrarytheROC-AUCsforalltheotherparameterswere
notsignificant.
Incervicalcancergroup,onlytheROC-AUCfortheDVwas
statis-ticallysignificantforbothGandriskclassification(Tables7and8).
TheROC-AUCsforalltheotherparameterswerenotsignificantfor
theriskclassification.OnthecontrarytheROC-AUCsforADCratio
(methodAandC)andT2volumewerestatisticallysignificantforG.
Althoughnosignificantdifferencewasobservedwhencompared
theROC-AUCsforDV,T2volumeandADCratio,ROC-AUCforDV
showedahighervaluethantheothers.
4. Discussion
Quantitative imaging characterization using advanced
tech-niquesiscurrentlyanongoingtopicinoncology.Theresultsofthis
studysuggestthatDVshowslowintra-/inter-observervariability
andcorrelateswithbothgradingandriskclassificationofcervical
andendometrialcancers.TheDVvaluesweresignificantlydifferent
amongGandriskgroupsinbothendometrialandcervicalcancers.
Moreover,theDVperformsbetterthanmorecommonlyusedADC
values,asshownbytheROCanalysis.
Thepotentialroleof ADCvalues inthepreoperative
evalua-tionofGandriskclassificationofcervicalandendometrialcancer
iscontroversial[7,10,12–14].Theheterogeneousreportedresults
maybeinpartexplainedbyseveraldifferentissues:theADCmin
valuereflectsthefunctionalactivityofonlyasmallsampleofthe
totalongoingneoplasticprocess;theADCmeancalculationmaybe
influencedbytheROIpositioningandsize;theADCratiodepends
ontheaccuracyofbothADCminandADCmeanvalues.Therefore,
thesecontroversialresultssuggestthenecessitytofindan
alter-nativemethodtoanalyzetheADCmaps.Inthepresentstudy,the
DVovercomesthelimitofeachADCvalueandeachmethodofADC
measurementintheassessmentofgradingandriskclassificationof
bothcervicalandendometrialcancers.TheDVselectstheportions
ofthetumorwithhighcellulardensities,whichrepresenttheactive
tumorburdenand indicatetheaggressivenessof theneoplasm.
Thisobservationiscorroboratedbythefollowingpreviousfindings
aboutPET/CTandPET/MRI:(a)theMTVrepresentsmoreaccurately
thanthemaximumstandardizeduptakevalue(SUVmax)theactive
Fig.6. a,b,c,dIntra-(aandb)andinter-(candd)observersreproducibilityfordiffusionvolume(DV)measurements(expressedincc)inaxial(aandc)andsagittalplanes (bandd)forendometrialcancer.Fortheintra-observerreproducibility,thesameoperatorcalculatedtwicetheDVinbothaxialandsagittalplanes:axial1DV,axial2DV, sagittal1DVandsagittal2DVofthesameoperatorwereobtained.Fortheinter-observersreproducibility,asecondoperatorcalculatedtheDVinbothaxialandsagittal planes:axial2DVandsagittal2DVofthesecondoperatorwereobtained.Bland–AltmanplotsoftheaverageoftheaxialDV(xaxis)againstthedifferencebetweenaxial DV(yaxis)ofthetwomeasurementsperformedbythesameoperator(a)andbythetwooperators(c);thecontinuouslinerepresentsthemeanabsolutedifferenceinDV betweenthetwomeasures;thedashedlinesrepresentthe95%confidenceintervalsofthemeandifferences.Bland–Altmanplotsoftheintra-observerreproducibility(b) andinter-observerreproducibility(d)forsagittalplane.
Table5
ROC-AUCsrelativetoGclassificationofendometrialcancer.
AUC SE 95%C.I. pValue
ADCmean(methodA) 0.609 0.124 0.407–0.786 0.3788 ADCmean(methodB) 0.636 0.127 0.434–0.808 0.2824 ADCmean(methodC) 0.575 0.131 0.375–0.758 0.5688 ADCmin(methodA) 0.599 0.121 0.398–0.778 0.4163 ADCmin(methodB) 0.653 0.124 0.451–0.822 0.2158 ADCmin(methodC) 0.650 0.132 0.447–0.819 0.2554 ADCratio(methodA) 0.541 0.108 0.343–0.729 0.7050 ADCratio(methodB) 0.531 0.138 0.334–0.720 0.8245 ADCratio(methodC) 0.619 0.134 0.417–0.795 0.3730 T2volume 0.683 0.127 0.468–0.852 0.1521 DV 0.762 0.106 0.564–0.901 0.0137*
AUC:Areaunderthecurve;SE:Standarderror;C.I.:Confidenceinterval.
*Apvalue<0.05wasconsideredsignificant.
metabolicactivityonFDG-PETalsohavemorerestricteddiffusion
onDWI,indicatinggreatercelldensity[30–31].
Since both endometrial and cervical cancers show ADC
val-uessignificantlylowerthannormalendometrial/cervicaltissueor
benignlesions[6,7,10–15]aswellastheviablesolidneoplastic
tis-suerespecttotumornecrosis[32],wesetanasymmetricwindow
aroundtheADCmeanshiftedtowardsthelowestvaluestomake
Fig.7.a,b,c,dIntra-(aandb)andinter-(candd)observersreproducibilityforthediffusionvolume(DV)measurementsexpressedinccinaxial(aandc)andsagittalplanes (bandd)forthecervicalcancer.Fortheintra-observerreproducibility,thesameoperatorcalculatedtwicetheDVinbothaxialandsagittalplanes:axial1DV,axial2DV, sagittal1DVandsagittal2DVofthesameoperatorwereobtained.Fortheinter-observersreproducibility,asecondoperatorcalculatedtheDVinbothaxialandsagittal planes:axial2DVandsagittal2DVofthesecondoperatorwereobtained.Bland–AltmanplotsoftheaverageoftheaxialDV(xaxis)againstthedifferencebetweenaxial DV(yaxis)ofthetwomeasurementsperformedbythesameoperator(a)andbythetwooperators(c);thecontinuouslinerepresentsthemeanabsolutedifferenceinDV betweenthetwomeasures;thedashedlinesrepresentthe95%confidenceintervalsofthemeandifferences.Bland–Altmanplotsoftheintra-observerreproducibility(b) andinter-observerreproducibility(d)forsagittalplane.
Table6
ROC-AUCsrelativetoriskclassificationofendometrialcancer.
AUC SE 95%C.I. pValue
ADCmean(methodA) 0.541 0.120 0.343–0.729 0.7342 ADCmean(methodB) 0.602 0.123 0.401–0.781 0.4051 ADCmean(methodC) 0.558 0.131 0.359–0.74 0.6589 ADCmin(methodA) 0.514 0.118 0.319–0.706 0.9085 ADCmin(methodB) 0.626 0.121 0.424–0.800 0.2968 ADCmin(methodC) 0.622 0.129 0.421–0.797 0.3413 ADCratio(methodA) 0.650 0.103 0.447–0.819 0.1445 ADCratio(methodB) 0.639 0.118 0.437–0.811 0.2358 ADCratio(methodC) 0.585 0.130 0.385–0.767 0.5129 T2volume 0.714 0.128 0.500–0.875 0.0929 DV 0.816 0.105 0.625–0.936 0.0025*
AUC:Areaunderthecurve;SE:Standarderror;C.I.:Confidenceinterval.
* Apvalue<0.05wasconsideredsignificant.
burden.Asstatedinmaterialandmethodssection,thenecessity
tofixtheHT(“meanvalueplusoneSD”)wasintendedtoexclude
theinactiveregionsofthetumorsuchasinflammatory changes
accompanyingthetumoraswellasthecystic/necroticcomponents
notidentifiedbyvisualcomparisonwithunenhancedand
contrast-enhancedMRimagesandtopreventtheerroneousinclusionof
contiguousnormaltissues,whiletheLT(“meanvalueminusthree
Table7
ROC-AUCsrelativetoGclassificationofcervicalcancer.
AUC SE 95%C.I. pValue
ADCmean(methodA) 0.595 0.188 0.382–0.785 0.6126 ADCmean(methodB) 0.680 0.163 0.465–0.851 0.2706 ADCmean(methodC) 0.620 0.166 0.406–0.805 0.4692 ADCmin(methodA) 0.520 0.200 0.313–0.722 0.9205 ADCmin(methodB) 0.575 0.159 0.363–0.768 0.6381 ADCmin(methodC) 0.675 0.132 0.460–0.847 0.1833 ADCratio(methodA) 0.850 0.111 0.651–0.960 0.0015* ADCratio(methodB) 0.520 0.159 0.313–0.722 0.8996 ADCratio(methodC) 0.850 0.111 0.651–0.960 0.0016* T2volume 0.825 0.0865 0.605–0.952 0.0002* DV 0.895 0.0728 0.707–0.981 <0.0001*
AUC:Areaunderthecurve;SE:Standarderror;C.I.:Confidenceinterval.
*Apvalue<0.05wasconsideredsignificant.
Table8
ROC-AUCsrelativetoriskclassificationofcervicalcancer.
AUC SE 95%C.I. pValue
ADCmean(methodA) 0.610 0.117 0.397–0.797 0.3464 ADCmean(methodB) 0.685 0.108 0.470–0.854 0.0868 ADCmean(methodC) 0.513 0.119 0.307–0.716 0.9134 ADCmin(methodA) 0.503 0.120 0.298–0.707 0.9784 ADCmin(methodB) 0.558 0.121 0.348–0.755 0.6292 ADCmin(methodC) 0.513 0.121 0.307–0.716 0.9147 ADCratio(methodA) 0.692 0.108 0.477–0.859 0.0749 ADCratio(methodB) 0.565 0.119 0.354–0.760 0.5864 ADCratio(methodC) 0.513 0.120 0.307–0.716 0.9136 T2volume 0.839 0.102 0.621–0.959 0.0929 DV 0.948 0.0403 0.779–0.997 <0.0001*
AUC:Areaunderthecurve;SE:Standarderror;C.I.:Confidenceinterval.
*Apvalue<0.05wasconsideredsignificant.
besubjecttonoiseorartifactscontamination.Relativetothislast
instance,verylow(nonbiologicallymeaningful)ADCvaluesmaybe
observedalsoinregionswithlowsignalinb0imagesinwhichthe
voxelsareaffectedbyrandomsignallossduetonoiseorartifacts.
Inthesevoxels,ADCvaluescanbeverylowconsideringthatthe
ADCcalculationisbasedontheratiobetweenthesignalmeasured
ontheb0andtheb1images.
Everymethodofsegmentationrequeststhechoiceofa
thresh-oldorarangeandthischoicemayrepresentasourceofbias.For
thisreasonthethresholdortherangehastobechosenonthebasis
ofreasonablecriteria.Moreover,everymethodof segmentation
maybeimproved.TheexampleofMTVisthatdifferentmethodsof
segmentationhavebeenproposedasfixed(absoluteSUVmax,
per-centageofSUVmax)orgradientthresholds[33–36].Thatmeans
thattheMTVis stilllargelyusedasbiologicalimaging
parame-ter,althoughthemethodtocalculateithaschangedinthecourse
oftime.Someobservationshavetobereportedonthethresholds
usedinourstudy:(1)testingthreedifferentcoupleofthresholds
(“ADCmeanminus3/plus1SD”;“ADCmeanminus2,5/plus1,5SD”;
“ADCmeanminus2/plus2SD”)in8patients,weobservedthatthe
range“ADCmeanminus3/plus1SD”recruitedfewvoxelslocated
outsidethesolidpartofthetumorinonlyapatient,withan
over-allbetterperformancethantheothersranges;settingtherange
“ADCmeanminus3/plus1SD”,themanuallyeditingofthe
auto-matedsegmentationdidnotrequestrobustcorrectionsinanyofthe
53patients;(2)therange“ADCmeanminus3/plus1SD”isbased
onreasonableandrestrictivethresholdsfixedtoincludehigh
cel-lulardensityvoxels;(3)therange“ADCmeanminus3/plus1SD”
offeredanexcellentcorrelation(ICC>0.81)betweenthe
measure-mentsobtainedontheaxialandthesagittalplanes,anexcellent
agreement(ICC>0.81)attheintra-andinter-observervariability,
asignificantdiscriminationofthegradingandtherisk
classifica-tionofbothcervicalandendometrialcancers;(4)theDVappears
asavalidbiologicalimagingparameter.Ofcourse,themethodof
segmentationweproposeinthepresentstudymaybeimproved
andrefinedwithincreasingdatacollectionandanalysis.
ThevolumesexpressedbytheDVweredifferentfromthose
computedonT2imagesandcorrelated significantlywiththeG
andtheriskclassification.ThisfindingsuggeststhattheDVand
themorphologicalvolume representdifferentaspectsofa
neo-plasmandthattheformerisnotexclusivelydefinedbytheguide
ofanatomicalimages.Apossibleexplanationofthisresultisthat
theinflammatorychangesaccompanyingtheneoplasticinfiltration
givesanoverestimationoftumorsizeonT2 images.The
asym-metric window around theADCmeanused tocalculate theDV
mayallowexcludingalsotheinflammatoryperilesionalreaction.
AsimilardiscrepancywasobservedbetweenFDG-PETvolume(at
SUVmaxcut-offpercentageof40%)andT2-weightedvolumein
cer-vicalcancerwithPET/MRI[37].Ontheotherhand,thesameAuthors
reportednosignificantdifferencebetweenFDG-PET(atSUVmax
cut-offpercentageof35%and40%),T2-weightedandDWIvolumes
incervicalcancerwithPET/MRI[31].Themaindifferencewithour
studyis relatedtothemethodof tumoroutlining:theAuthors
selectedthewholegrossneoplasticvolume,whilewelookedonly
atthesolidparts.Theseresultsaswellasourssuggestthat,rather
thandemonstratingoverlapordiscrepancybetween
morphologi-calandfunctionalvolume,itiscrucialtofindandemphasizewhich
morphologicalandfunctionalinformationmayberepresentative
ofthetumorbiologicalcharacteristics.
Recently,thequantitativehistogramanalysisofADCmapshas
beenproposedintheidentificationofadversehistological
char-acteristicsofstageIcervicalcarcinoma[18].Wedidnotconsider
usingthequantitativehistogramanalysisinourpopulationbecause
themethodlooksattheheterogeneityevaluatingthewhole
neo-plasm(solidcellular,inflammatory,cystic/necroticparts)rather
thantheactiveneoplasticburdenwhichrepresentsthegoalofour
val-ueswouldbeanywaypresentingettingDVfromthepercentilesor
theskeworthekurtosismeasurements.
ThevariationsinROIsizeandpositioninghavebeenshowedto
haveasignificanteffectontumorADCvaluesandinter-observer
variabilityinpatientswithrectalcancer[38].Moreover,itisnot
clearlydefinedwhichADCvalue(ADCmin,ADCmean,ADCratio)
ismorerepresentativeoftumoraggressiveness[6,7,10–15].Asa
consequence,wechosetoanalyzetheADCmapswiththree
differ-entsetsofROIsaswellastotestthreedifferentvaluesofADCto
preventthatthelackofcorrelationwithclinicalparameterscould
beattributedtothemethodofanalysis.Inourseries,the
correla-tionofallthetestedADCvalueswithgradingandriskclassification
resultedunsatisfyingforbothendometrialandcervicalcancer.On
thecontrary,theDVappearedtoberepresentativeofthe
aggres-sivenessofthetumor.
Inconclusion,theDVcorrelateswithgradingandrisk
classifi-cationofbothcervicalandendometrialcancersaswellasitshows
lowintra-/inter-observervariability.Prospectivestudiesinlarger
populationshavetobedesignedtoconfirmourobservations.
References
[1]T.Koyama,K.Tamai,K.Togashi,Stagingofcarcinomaoftheuterinecervix andendometrium,Eur.Radiol.17(8)(2007)2009–2019.
[2]G.Rechichi,S.Galimberti,M.Signorelli,P.Perego,M.G.Valsecchi,S.Sironi, Myometrialinvasioninendometrialcancer:diagnosticperformanceof diffusioneweightedMRimagingat1.5-T,Eur.Radiol.20(3)(2010)754–762.
[3]A.Andreano,G.Rechichi,P.Rebora,S.Sironi,M.G.Valsecchi,S.Galimberti,MR diffusionimagingforpreoperativestagingofmyometrialinvasioninpatients withendometrialcancer:asystematicreviewandmeta-analysis,Eur.Radiol 24(6)(2014)1327–1338.
[4]H.A.Rowley,P.E.Grant,T.P.Roberts,DiffusionMRimaging:theoryand applications,NeuroimagingClin.N.Am.9(2)(1999)343–361.
[5]D.M.Koh,D.J.Collins,Diffusion-weightedMRIinthebody:applicationsand challengesinoncology,AJRAm.J.Roentgenol.188(6)(2007)1622–1635.
[6]S.Naganawa,C.Sato,H.Kumada,T.Ishigaki,S.Miura,O.Takizawa,Apparent diffusioncoefficientincervicalcanceroftheuterus:comparisonwiththe normaluterinecervix,Eur.Radiol.15(1)(2005)71–78.
[7]P.Z.McVeigh,A.M.Syed,M.Milosevic,A.Fyles,M.A.Haiderd, Diffusion-weightedMRIincervicalcancer,Eur.Radiol.18(5)(2008) 1058–1064.
[8]E.Sala,A.Rockall,D.Rangarajanc,R.A.Kubik-Huch,Theroleofdynamic contrast-enhancedanddiffusionweightedmagneticresonanceimagingin thefemalepelvis,Eur.J.Radiol.76(3)(2010)367–385.
[9]S.Punwani,Diffusionweightedimagingoffemalepelviccancers:concepts andclinicalapplications,Eur.J.Radiol.78(1)(2011)21–29.
[10]G.Rechichi,S.Galimberti,M.Signorelli,C.T.Franzesi,P.Perego,M.G. Valsecchi,etal.,Endometrialcancer:correlationofapparentdiffusion coefficientwithtumorgrade,depthofmyometrialinvasion,andpresenceof lymphnodemetastases,AJRAm.J.Roentgenol.197(1)(2011)256–262.
[11]S.Fujii,E.Matsusue,J.Kigawa,S.Sato,Y.Kanasaki,J.Nakanishi,etal., Diagnosticaccuracyoftheapparentdiffusioncoefficientindifferentiating benignfrommalignantuterineendometrialcavitylesions:initialresults,Eur. Radiol.18(2)(2008)384–389.
[12]S.H.Shen,Y.Y.Chiou,J.H.Wang,M.S.Yen,R.C.Lee,C.R.Lai,etal.,
Diffusion-weightedsingle-shotecho-planarimagingwithparalleltechnique inassessmentofendometrialcancer,AJRAm.J.Roentgenol.190(2)(2008) 481–488.
[13]K.Tamai,T.Koyama,T.Saga,S.Umeoka,Y.Mikami,S.Fujii,etal.,Diffusion weightedMRimagingofuterineendometrialcancer,J.Magn.Reson.Imaging 26(3)(2007)682–687.
[14]N.Bharwani,M.E.Miquel,A.Sahdev,P.Narayanan,G.Malietzis,R.H.Reznek, etal.,Diffusion-weightedimagingintheassessmentoftumourgradein endometrialcancer,Br.J.Radiol.84(1007)(2011)997–1004.
[15]Y.Inada,M.Matsuki,G.Nakai,F.Tatsugami,M.Tanikake,I.Narabayashi,etal., BodydiffusionweightedMRimagingofuterineendometrialcancer:isit helpfulinthedetectionofcancerinnonenhancedMRimaging?Eur.J.Radiol. 70(1)(2009)122–127.
[16]K.Cao,M.Gao,Y.S.Sun,Y.L.Li,Y.Sun,Y.N.Gao,etal.,Apparentdiffusion coefficientofdiffusionweightedMRIinendometrialcarcinoma-Relationship withlocalinvasiveness,Eur.J.Radiol.81(8)(2012)1926–1930.
[17]P.Beddy,A.C.O’Neil,A.K.Yamamoto,H.C.Addley,C.Reinhold,E.Sala,FIGO stagingsystemforendometrialcancer:addedbenefitsofMRimaging, Radiographics32(1)(2012)241–254.
[18]K.Downey,S.F.Riches,V.A.Morgan,S.L.Giles,A.D.Attygalle,T.E.Ind,etal., RelationshipbetweenimagingbiomarkersofstageIcervicalcancerand poor-prognosishistologicfeatures:quantitativehistogramanalysisof diffusion-weightedMRimages,AJRAm.J.Roentgenol.200(2)(2013) 314–320.
[19]F.Kuang,J.Ren,Q.Zhong,F.Liyuan,Y.Huan,Z.Chen,Thevalueofapparent diffusioncoefficientintheassessmentofcervicalcancer,Eur.Radiol.23(4) (2013)1050–1058.
[20]R.Bos,J.J.vanDerHoeven,E.vanDerWall,P.vanDerGroep,P.J.vanDiest,E.F. Comans,etal.,Biologiccorrelatesof(18)fluorodeoxyglucoseuptakein humanbreastcancermeasuredbypositronemissiontomography,J.Clin. Oncol.20(2)(2002)379–387.
[21]T.Higashi,N.Tamaki,T.Torizuka,Y.Nakamoto,H.Sakahara,T.Kimura,etal., FDGuptake,GLUT-1,glucosetransporterandcellularityinhumanpancreatic tumors,J.Nucl.Med.39(10)(1998)1727–1735.
[22]K.Ito,T.Kato,T.Ohta,etal.,Fluorine-18fluoro-2-deoxyglucosepositron emissiontomographyinrecurrentrectalcancer:relationtotumoursizeand cellularity,Eur.J.Nucl.Med.23(10)(1996)1372–1377.
[23]R.Fonti,M.Larobina,S.DelVecchio,S.DeLuca,R.Fabbricini,L.Catalano,etal., Metabolictumorvolumeassessedby18F-FDGPET/CTforthepredictionof outcomeinpatientswithmultiplemyeloma,J.Nucl.Med.53(12)(2012) 1829–1835.
[24]N.Colombo,E.Preti,F.Landoni,S.Carinelli,A.Colombo,C.Marini,etal., Endometrialcancer:ESMOclinicalpracticeguidelinesfordiagnosis, treatmentandfollowup,Ann.Oncol.22(2011)35–39,Suppl7:vii.
[25]S.J.Freeman,A.M.Aly,M.Y.Kataoka,H.C.Addley,C.Reinhold,E.Sala,The revisedFIGOstagingsystemforuterinemalignancies:implicationsforMR imaging,Radiographics32(6)(2012)1805–1827.
[26]N.Colombo,S.Carinelli,A.Colombo,C.Marini,D.Rollo,C.Sessa,etal., Cervicalcancer:ESMOClinicalPracticeGuidelinesfordiagnosis,treatment andfollow-up,Ann.Oncol.23(2012)27–32,Suppl7:vii.
[27]A.Berkowitz,S.Basu,S.Srinivas,S.Sankaran,S.Schuster,A.Alavi,
Determinationofwhole-bodymetabolicburdenasaquantitativemeasureof diseaseactivityinlymphoma:anovelapproachwith
fluorodeoxyglucose-PET,Nucl.Med.Commun.29(6)(2008)521–526.
[28]H.H.Chung,J.W.Kim,K.H.Han,J.S.Eo,K.W.Kang,N.H.Park,etal.,Prognostic valueofmetabolictumorvolumemeasuredbyFDG-PET/CTinpatientswith cervicalcancer,Gynecol.Oncol.120(2)(2011)270–274.
[29]G.Storto,E.Nicolai,M.Salvatore,[18F]FDG-PET-CTforearlymonitoringof tumorresponse:whenandwhy,Q.J.Nucl.Med.Mol.Imaging53(2)(2009) 167–180.
[30]J.R.Olsen,J.Esthappan,T.DeWees,V.R.Narra,F.Dehdashti,B.A.Siegel,etal., TumorvolumeandsubvolumeconcordancebetweenFDG-PET/CTand diffusion-weightedMRIforsquamouscellcarcinomaofthecervix,J.Magn. Reson.Imaging37(2)(2013)431–434.
[31]H.Sun,J.Xin,S.Zhang,Q.Guo,Y.Lu,W.Zhai,etal.,Anatomicalandfunctional volumeconcordancebetweenFDGPET,andT2anddiffusion-weightedMRI forcervicalcancer:ahybridPET/MRstudy,Eur.J.Nucl.Med.Mol.Imaging41 (5)(2014)898–905.
[32]M.Uhl,U.Saueressig,G.Koehler,U.Kontny,C.Niemeyer,W.Reichardt,etal., Evaluationoftumournecrosisduringchemotherapywithdiffusion-weighted MRimaging:preliminaryresultsinosteosarcomas,Pediatr.Radiol.36(12) (2006)1306–1311.
[33]M.Y.Choi,K.M.Lee,J.K.Chung,D.S.Lee,J.M.Jeong,J.G.Park,etal.,Correlation betweenserumCEAlevelandmetabolicvolumeasdeterminedbyFDGPETin postoperativepatientswithrecurrentcolorectalcancer,Ann.Nucl.Med.19 (2)(2005)123–129.
[34]E.H.Dibble,A.C.Alvarez,M.T.Truong,G.Mercier,E.F.Cook,R.M.
Subramaniam,18F-FDGmetabolictumorvolumeandtotalglycolyticactivity oforalcavityandoropharyngealsquamouscellcancer:addingvalueto clinicalstaging,J.Nucl.Med.53(5)(2012)709–715.
[35]P.Sridhar,G.Mercier,J.Tan,M.T.Truong,B.Daly,R.M.Subramaniam,FDGPET metabolictumorvolumesegmentationandpathologicvolumeofprimary humansolidtumors,AJRAm.J.Roentgenol.202(5)(2014)1114–1119.
[36]N.Withofs,C.Bernard,C.VanderRest,P.Martinive,M.Hatt,S.Jodogne,etal., FDGPET/CTforrectalcarcinomaradiotherapytreatmentplanning: comparisonoffunctionalvolumedelineationalgorithmsandclinical challenges,J.Appl.Clin.Med.Phys.15(5)(2014)4696.
[37]S.Zhang,J.Xin,Q.Guo,J.Ma,Q.Ma,H.Sun,Comparisonoftumorvolume betweenPETandMRIincervicalcancerwithhybridPET/MR,Int.J.Gynecol. Cancer24(4)(2014)744–750.
[38]D.M.Lambregts,G.L.Beets,M.Maas,L.Curvo-Semedo,A.G.Kessels,T. Thywissen,etal.,TumourADCmeasurementsinrectalcancer:effectofROI methodsonADCvaluesandinterobservervariability,Eur.Radiol.21(12) (2011)2567–2574.