Transcript
Page 1: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

NCI Quan)ta)ve Imaging Network (QIN)

Opportuni)es for QI Tools in Breast Oncology

For QIN: Nola Hylton, UCSF

Maryellen Giger, University of Chicago

COI:M.L.G.isastockholderinR2/Hologic,co-founderandequityholderinQuanCtaCveInsights,shareholderinQview,andreceivesroyalCesfromHologic,GEMedicalSystems,MEDIANTechnologies,RiverainMedical,Mitsubishi,andToshiba.ItistheUniversityofChicagoConflictofInterestPolicythatinvesCgatorsdisclosepubliclyactualorpotenCalsignificantfinancialinterestthatwouldreasonablyappeartobedirectlyandsignificantlyaffectedbytheresearchacCviCes.

Page 2: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

The Quan)ta)ve Imaging Network (QIN)

TheQINisanNCIProgramjointiniCaCvetobringquanCtaCveimagingmethodsintoclinicaluClityformeasuringresponsetotreatmentandsupporCngclinicaldecision-making25teamsintheQINfocusonimprovingquanCtaCveresultsfromclinicalimagesforaspecificcancerproblemCross-NetworkWorkingGroupsaddress:1)ImageAnalysisandPerformanceMetrics(MRIandPET/CTSubgroups);2)BioinformaCcs/ITandData-sharing;3)ClinicalTrialsDesignandDevelopment

Page 3: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

3

Networkintent:buildconsensus,sharedataandtools.

The Quan)ta)ve Imaging Network

•  25acCveteams(twofundedthroughtheCanadianGovernment)•  12associatemembersfromUSand7foreigncountries

•  Over46toolsunderdevelopmentandvalidaCon

Page 4: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Quan)ta)ve Imaging

QuanCtaCveimagingistheextracConofquanCfiable(measurable)featuresfrommedicalimagesfortheassessmentofnormalortheseverity,degreeofchange,orstatusofadisease,injury,orchroniccondiConrelaCvetonormalItisthecombinaConofimaging,analyCcalmethods,andinformaCcs

Page 5: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

PaCentStatus

ImageAcquisiCon

PaCentImages

AlgorithmProcessing

QuanCtaCveResults

ClinicalDecision

QIN: From Informa)on to Knowledge

Page 6: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

The Complexi)es of Quan)ta)ve Tools in Clinical Trials

•  Clinicalchallenge:–  IdenCfytheclinicallymostmeaningfulimagingmarkerforthestudyobjecCve

•  Technicalchallenges:–  ImagestandardizaCon–  ImageacquisiCon– Datatransferand/oranalysis–  SiteversuscentralquanCtaCveanalysis–  ImageanalysistooldistribuConandvalidaCon

Page 7: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Representa)ve tools developed by QIN teams Tool Modality Purpose

LymphnodesegmentaCon MRI LymphnodesegmentaCon

HologicAegisSER MRI VolumetricbreasttumorsegmentaConQuanCtaCveInsightsQuantX MRI VolumetrictumorsegmentaConandmachinelearning

diagnosCcs

Xcal PET MulCcenterPETSUVcross-calibraCon

AutoPERCIST PET PERCISTresponseanalysisforFDG-PET

LungSegmentaCon CT VolumetriclungnodulesegmentaCon

Radiomicsanalysis CT Lung,headandneckradiomicsanalysis

MassesCmaCon CT MusclemassofcancerpaCents

ePAD Imageanalysis ImageannotaConandquanCtaCveanalysis

Slicer Imageanalysis Imageanalysisandsurgicalplanning

Page 8: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Why we (the QIN) are here

•  ToidenCfyoncologytrialswherequanCtaCveimagingbiomarkersandQINtoolscansupportoutcomesbyimprovingefficacy,efficiency,orstudypower

• QIN-NCTNPlanningmeeCngrecommendaCons(December2016):

•  QINtoolintegra.onintoclinicaltrialsshouldstartasearlyaspossibleintrialdevelopment

•  Increaseddialogueneededbetweenimagersandoncologists

•  Presenta.onsbyQINmembersat(1)theAlliancePlenarysession,(2)selecteddiseasesitecommiAees,and(3)ImagingcommiAee

Page 9: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Breast Quan)ta)ve Imaging in Clinical Trials

Page 10: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

•  Func.onaltumorvolume(FTV)predictsrecurrence-freesurvival(RFS)

•  FTVisastrongerpredictorofRFSthanpathologiccompleteresponse(pCR)

•  FTVpredictsRFSasearlyasaKer1cycleofstandardanthracycline-basedchemotherapy

MRIatbaseline,1cycle,betweenACandT,andpre-surgery

Hylton et al., RADIOLOGY 2015

I-SPY 1: ACRIN 6657 & CALGB 150007– Contrast-enhanced MRI for assessing breast cancer response to neoadjuvant chemotherapy

Page 11: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

•  FTVpredic+veperformanceandop+malmeasurement+mepointdifferbybreastcancersubtype

Hylton et al., RADIOLOGY 2015

Early-treatment

Pre-surgery

HR+/HER2- HER2+ HR-/HER2- (TN)

I-SPY 1: ACRIN 6657 & CALGB 150007– Contrast-enhanced MRI for assessing breast cancer response to neoadjuvant chemotherapy

Page 12: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

•  Drugs“graduate”fromI-SPY2whentheyreachaBayesianpredicCveprobabilityofachieving80%successinasubsequentphaseIIIstudy

•  Drugsgraduatewithinsubtypesdefinedbyhormonereceptor(HR)status,HER2statusandMammaprintscore

•  DrugscanbedroppedforfuClity•  I-SPY2isanadap.vely-randomizedphaseIItrialtes.ng

novelagentsforbreastcancer•  IncorporatesMRItumorvolumeinthepa.ent

randomiza.onalgorithm

>2150pa.entsregistered;>1220randomized;>1070withsurgeryasofOct20176druggraduatestodate

I-SPY 2 breast cancer trial

Page 13: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

•  ACRIN6698:sub-studyofI-SPY2tesCngdiffusion-weightedMRI(10sites)•  406I-SPY2paCentsenrolled;272ontreatmentcombinedforanalysis•  DWIaddedtostandardDCE-MRI•  Apparentdiffusioncoefficient(ADC)measuredusingDWI•  Preliminaryresults(presentedatASCO2017):

Ø  ADCandchangeinADCatmid-therapyandpre-surgerypredictpCRØ  Variablepredic+onbysubtype,highestinHR+/HER2-

MulC-focalinvasiveductalcarcinoma.Pre-treatmentDCEMRI1(leo)andDWIb800(right)

DWImeasurestherandommo.onofwaterin.ssueProvidesinforma.onaboutcelldensityandmicrostructure

DCE DWI

ACRIN 6698 - Breast diffusion-weighted MRI (DWI) to predict response to neoadjuvant chemotherapy

Page 14: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

•  Endpoints:•  Primary:radiographicresponseletrozoleonMRI

•  ChangeinMRItumorvolume

•  Secondary:•  Mammographicextentofdisease

•  CandidacyforbreastconservaCon•  Frequencyofre-excisions•  PathCR•  Invasivecanceratexcision

CALGB 40903: Phase II Single-Arm Study of Neoadjuvant letrozole for ER(+) postmenopausal DCIS (PI: Shelley Hwang)

Page 15: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

ACRIN 6688: FLT PET to Measure Early Breast Cancer Response (PI: Lale Kostakoglu)

Best ΔSUVmax cut-off for predicting pCR = -51% (sensitivity 56%;specificity 79%).

Pre-Therapy

7 d Post-

(Kostakoglu, J Nucl Med, 2015) ACRIIN and U Wash QIN U01

Page 16: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Next Steps: Benefit of using exis)ng and future Clinical Trial data Increase effec)veness & efficiency Incorporate automated, objec)ve computer-extracted biomarkers (radiomics) & develop decision tools using machine learning. Enable efforts to standardize, verify quality, and validate with exis)ng and future Clinical Trial data.

Page 17: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Incorpora)ng automated computer-extracted characteris)cs (radiomics) into response assessment

(METV on ACRIN 6657 data: only pre-treatment & early treatment imaging exams)

EsCmatedKaplan-Meierrecurrence-freesurvivalesCmatesforMETVNIHQINGrantU01CA195564

Page 18: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

4D DCE MRI images

Computer-Extracted Image Phenotypes

Size

Shape

Morphology

Contrast Enhancement

Texture

Curve

Variance

……

Computerized Tumor Segmentation

Radiologist-indicated Tumor Center

CADpipeline=radiomicspipeline

InputtoClassifier(LDA,SVM)NIHQINGrantU01CA195564

Incorporating machine learning into assessing diagnosis, molecular classification, & response assessment

Computer-extraction of biomarkers (features) followed by training of predictive classifiers

Page 19: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

MulC-insCtuConal,MulC-disciplinaryCollaboraCon

cancerimagingarchive.netBreast Cancer cases

Clinical/Histopathology/GenomicdatadownloadedbyTCGAAssembler&Molecularsubtyping/riskofrecurrence

valuesbyPerouLab

TumorlocaCononMRIdeterminedbyconsensusofthreeoftheTCIAradiologists

MRIsof91cases(GE1.5T)collectedbyTCIA

MRIsof91casesdownloadedtoUChicagoforcomputaConalMRItumor

phenotyping(radiomics)

cancergenome.nih.gov

Page 20: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

FromtheTCIARadiomics--EnhancementTextureofTumorHeterogeneityappearsPredicCveofMolecularSubtype–ClinicalPrognos8cValue

4 55

10 5 10

Kendall test results for trends; p-value=0.0055

LiH,ZhuY,BurnsideES,….PerouCM,JiY,GigerML:QuanCtaCveMRIradiomicsinthepredicConofmolecularclassificaConsofbreastcancersubtypesintheTCGA/TCIADataset.npjBreastCancer(2016)2,16012;doi:10.1038/npjbcancer.2016.12;publishedonline11May2016.

Page 21: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Good Prognosis Case (left)

Poor Prognosis Case (right)

Cancer Subtype Luminal A Basal-like OncotypeDX Range [0, 100]

14.4 (low risk of breast cancer

recurrence)

100 (high risk of breast cancer

recurrence) MammaPrint

Range [0.848, -0.748] 0.67

(good prognosis) -0.54

(poor prognosis) PAM50 ROR-S (Subtype)

Range [-7.42, 71.76] -2.2

(low risk of breast cancer recurrence)

56.3 (high risk of breast cancer

recurrence) PAM50 ROR-P

(Subtype+Proliferation) Range [-13.21, 72.38]

0.96 (low risk of breast cancer

recurrence)

53.2 (high risk of breast cancer

recurrence) MRI Tumor Size

(Effective Diameter) Range [7.8 54.0]

16.8 mm

21.7 mm

MRI Tumor Irregularity Range [0.40 0.84]

0.438

0.592

MRI Tumor Heterogeneity (Entropy)

Range [6.00 6.59]

6.27

6.51

!

Multi-gene assays of risk of recurrence

Radiomics for “virtual” biopsy

Clinical Therapeutic Response Assessment Value

LiH,ZhuY,BurnsideES,….PerouCM,JiY*,GigerML*:MRIradiomicssignaturesforpredicCngtheriskofbreastcancerrecurrenceasgivenbyresearchversionsofgeneassaysofMammaPrint,OncotypeDX,andPAM50.RadiologyDOI:hwp://dx.doi.org/10.1148/radiol.2016152110,2016.

Page 22: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

IMAGINGGENOMICS–USINGVIRTUALBIOPSIESPATHWAYTRANSCRIPTIONALACTIVITIESASSOCIATEDWITHMRIQUANTITATIVEFEATURES

ZhuY,LiH,…GigerML*,JiY*:DecipheringgenomicunderpinningsofquanCtaCveMRI-basedradiomicphenotypesofinvasivebreastcarcinoma.Nature–ScienCficReports5:17787(2015)

ShapeSize

Heterogeneity

Page 23: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

Opportuni)es for NCTN-QIN Collabora)ons

1.  QINcanprovideexperCsetoguideimagingneedsforNCTNtrials•  QINinvesCgatorsareeagertoparCcipateinNCTNtrials

2.  QINinvesCgatorsseekopportuniCestoaddexploratorybiomarkerstoNCTNtrials,ooenwithoutaddedcost•  QINteamarefundedtodevelopQItools,andrelishthechancetotesttoolsprospecCvelyintrials

•  AddimagingtranslaConalsciencetoNCTNtrials

3.  EnhancedpartnershipforoncologyandimaginginvesCgatorsinNCTNtrials•  Commongoalsofimprovedthequalityandefficiencyofcancerclinicaltrials

Page 24: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

QINprogramofficeRobertNordstrom [email protected],CancerImagingProgram

LoriHenderson [email protected],ClinicalTrialsBranch,CancerImagingProgram

QIN Contact Information

Page 25: NCI Quan)ta)ve Imaging Network (QIN) · NCI Quan)ta)ve Imaging Network (QIN) Opportuni)es for QI Tools in Breast Oncology For QIN: Nola Hylton, UCSF Maryellen Giger, University of

QIN Presenta)ons at Alliance Annual Mee)ng

CommiMee QINRepresenta8veBreast NolaHyltonandMaryellenGigerExperimentalTherapeuCcs PaulKinahanandAmitaDaveGI LarrySchwartzandHugoAertsGU MichaelJacobsandAndryFedorovLymphoma RichWahlandDaveMankoffNeuro-Oncology MichaelKnoppandJayshareeKalpathyRadiaCon-Oncology Hui-KuoShuandYueCaoRespiratory JohnBuayandMichaelMcNiw-GrayIROC Xiao,Rosen,Knopp,andFitzgerald