Download Developing novel image fusion registration for targeted biopsies for

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Prostate-specific antigen wikipedia , lookup

Transcript
Developingnovelimagefusionregistrationfortargetedbiopsiesfor
earlyprostatecancerdetection
LeadInvestigator:
VincentGnanapragasam
AcademicUrologyGroup,DepartmentofSurgeryandOncology
http://www.cambridgecancercentre.org.uk/users/vjg29
CoInvestigators:
Carola-BibianeSchoenlieb
DepartmentofAppliedMathematicsandTheoreticalPhysics
http://www.damtp.cam.ac.uk/user/cbs31/Home.html ProfessorJAston
ProfessorofStatistics,StatisticalLaboratory
DrGrahamTreece
ReaderinInformationEngineering,DepartmentofEngineering
This PhD Studentship within the Urological Malignancies Programme offers a highly unique
opportunity to work across a clinical, mathematical and engineering interface, and bring together
three key disciplines to address a current conundrum in clinical practice and patient safety. The
strong background of each of the co-supervisors in their respective disciplines in addressing the
different aspects of the project will ensure its success and provide excellent supervision to the
student.Inaddition,theprojecthasstrongpotentialtobeextendedfurthertofuturestudies,and
rapidclinicalapplicationforpatientbenefit.
Thisprojectmightappealtostudentswithaninterestinappliedmathematics,statistics,urologyand
computational imaging. A strong background in one of these subjects as well as programming
experienceisessential.
ProjectDescription
Background
ProstatecanceristheprevalentmalecancerintheWesternworldanditsincidenceisrisingrapidly
(http://www.cancerresearchuk.org/about-cancer/type/prostate-cancer/). The most common
methodofdiagnosingprostatecanceriswithaneedlebiopsyoftheprostateguidedbyatransrectal
ultrasound (US) probe inserted into the rectum (https://www.nice.org.uk/guidance/). It is now
known however that over 30% of cancers will be missed using this method because ultrasound
cannotvisualisewherealesionisintheprostate(Nelson,2014).
Pre-biopsyimagingisthereforefastbecomingastandardofcareinplanningthebiopsyapproachto
reduce the risk of missing cancers. Currently, pre-biopsy imaging is done by a magnetic resonance
imaging(MRI)scanoftheprostatewhichisthenusedtomapontoanultrasoundpicturetoguide
thebiopsy.Thiscanbedoneusingcognitive(orvisual)estimationofanMRIlesionandestimated
correlationwithanUSimage(performedbytheoperator)orbyimagefusiontechniques(whichis
moreaccurate).ThemostcommonmethodofimagefusionistosuperimposeanMRIimageontoan
ultrasoundpictureinaformofstaticfusion(Nienke,2016).Pre-markedareasintheprostateonthe
MRI representing tumour are then assumed to be concurrent with the areas of ultrasound and
samplesaretakenaccordingly.Thismethod,however,requiresboththepatientandtheultrasound
probetobeverystill(necessitatingageneralanaesthetictokeepthepatientstationary)aswellas
expensiveequipmentforthebiopsyguidingmethod(typicallyafixedgantryandstraightlineneedle
insertion).Analternativemethodistousemagneticfieldgeneratingsoftware(e.g.Uronavsystem)
todetectwhereabiopsyprobemightbeinrelationtotheprostate,andtousethistofuseMRIand
USimages.Thisagainrequiresinconvenientandexpensiveequipmentandisnotsuitableasadayto
daytoolfortheover80,000prostatebiopsiescarriedoutannuallyintheUK(1millionintheUSand
Europe).
Currentmethods,althougheffective,arethereforenotsuitableforgeneraluseacrossthethousands
ofunitsthatundertakeprostatebiopsies.Acriticalneedisasimplifiedtoolthatcanperformreal
time image fusion without the use of specific hardware or expensive clinical equipment. In this
proposal, we are seeking to address this by using computation and mathematical modelling to
develop novel anatomical registration techniques for real time image fusion guided biopsies and
which can be delivered as proprietary software. The student (who will be based in the Centre for
Mathematical Sciences) appointed to this project will benefit from collaborative working and joint
supervisioninurologicalsurgery(VJG),computationalscience/mathematics(CS/JA)andengineering
(GT) to deliver a clinic-ready application. Engineering already has a strong history of producing
clinically useful solutions and software (Gee, 2003). This project will also strongly complement
existingworklookingatnovelbiopsytechniques.
Hypothesis
Mathematicalandcomputationalmodellingtechniquescanbeusedtoaddresstheneedof
anatomicalbasedreal-timeimageregistrationforguidedprostatebiopsies.
Objectives
1.Tousereal-timeclinicalimagingfrompatientsundergoingbiopsiestoexploreanatomicalor
fiducialco-registrationtechniqueswithMRIandultrasoundimages.
2.Todevelopandtestnovelregistrationalgorithmstodeterminetheoptimummethodofimage
fusion.
3.Todevelopsoftwarethatwillallowreal-timeacquisitionofimagesandpresentableinaformat
thataclinicianundertakingprostatebiopsiescanusetoguidebiopsies,withouttheuseof
extrinsichardwareorsensor-guidingtechnology.
Proposedmethods
Image registration is the task of aligning two images. In the present case we are aiming for a
registrationthatalignsapre-recordedstaticMRimagewithframesofadynamicUSimagesequence
recorded during the biopsy. For developing a clinically practical image-guided prostate biopsy
pipeline,atailor-maderegistrationframeworkisnecessary,whichistunedtowardsfindingmutual
structuresinMRIandUSimagesthatcanbeusedforaligningtheminacomputationallycheapway.
Toaddressthis,thestudentwillundertakeresearchcoveringthefollowingthemes:
Theme 1: Novel US/MRI co-registration (Yr 1-2) (VJG/GT/CS): The formalisation and automated
detectionofmutualfeaturesofMRandUSimagesarelikelytorequiresophisticatedmathematical
concepts and clinical insight. In particular, the student will be expected to go beyond standard
intensity-based registration towards more complex local- and non-local feature-based registration
(Modersitzki, 2009, Lee 2015, Lotz 2015, Maass J 2015), especially tuned towards the properties
encodedinMRandUSimages.Toachievethis,thestudentwillusedatageneratedfromaclinical
series of men undergoing biopsies and who will have MRI and US images taken at procedure
(throughtheclinicalcollaborationwiththeurologydepartment).Usingthese,thestudentwilltest
anddevelopnewalgorithmsandapproachestoimageco-registration.
Theme 2: Rapid real-time computing registration (Yr 2-3) (CS/JA/GT): The registration needs to be
computable in real-time and on low-level hardware. However sophistication in the registration
techniqueusuallygoestogetherwithincreaseincomputationalcosts.Therefore,anotherchallenge
will be to build a comprehensive, yet computationally manageable registration framework by
enhanced engineering of derived mathematical principles. In particular, we will aim for low-level
imageregistrationwhichlinksthemostcharacteristicmutualfeaturesinbothimagestoeachother,
whileneglectingsomeofthecomputationallycostlyaspects.Here,computationaltimeneedstobe
balanced with registration accuracy in a reliable and statistically controllable manner. This step
requires engineering as well as statistical and mathematical expertise (collaboration between
engineeringandmathematics).
Theme 3: Translation to clinical application (Yr 3) (VJG/GT/CS): A unique aspect and deliverable of
this project is that the outputs from the work above will be developed into a tool that can be
simulated in the clinic to test potential application (i.e. the ability to see where a tumour is on a
fusedrealtimeimageanddirectabiopsyneedletothislesion).Todothisthestudentwillbegiven
anhonorarycontracttoworkalongsidetheurologyteam.Softwaretoolsdevelopedfromcombining
themes1and2abovewillbeusedinparallelwithstandardimagefusionsoftware(e.g.e.g.Biopsee
fusion system) already being used in clinics. Model simulated biopsies from areas identified using
thesoftwarewillbecomparedtoyieldsfromtruebiopsiestotestifthetoolcanaccuratelyidentify
areas of interest. This will also permit an iterative improvement in the design and performance of
thetool.
Thetacklingofallthreechallengesaboverequiresacontinuousfeedback-loopcheckingthevalidity
ofthehypothesislaidoutanditsclinicalrealisation.Assuchthisprojectwillprovideauniquenexus
ofmathematical,engineeringandclinicalsupervisiontosupportthestudenttoachievethegoalsof
thisproject
References
NelsonA(2014)Repeatprostatebiopsystrategiesafterinitialnegativebiopsy:meta-regression
comparingcancerdetectionoftransperineal,transrectalsaturationandMRIguidedbiopsy.PLoS
One.2013;8(2):e57480
A.H.Gee,R.W.Prager,G.M.TreeceandL.Berman,(2003),Engineeringafreehand3Dultrasound
system.PatternRecognitionLetters.24(4-5).
Modersitzki,J.(2009).FAIR:flexiblealgorithmsforimageregistration(Vol.6).SIAM.
J.Lee,X.Cai,C.-B.Schönlieb,andD.Coomes,(2015),Non-parametricImageRegistrationofAirborne
LiDAR,HyperspectralandPhotographicImageryofWoodedLandscapes,GeoscienceandRemote
Sensing,IEEETransactionson,53(11),6073-6084.
Lotz,J.,JanineOlesch,BenediktMuller,ThomasPolzin,PatriciaGaluschka,JudithLotz,Stefan
Heldmannetal.(2015)"Patch-BasedNonlinearImageRegistrationforGigapixelWholeSlide
Images.".
Maas,J.,Rumpf,M.,Schönlieb,C.,&Simon,S.(2015).Ageneralizedmodelforoptimaltransportof
imagesincludingdissipationanddensitymodulation.ESAIM:MathematicalModellingandNumerical
Analysis,49(6),1745-1769.
Applications
Toapplyforthisstudentshippleaseseehttp://www.cambridgecancercentre.org.uk/studentships
ForgeneralenquiriespleasecontactTinaThorn [email protected]
Forfurtherinformationorquestionsrelatingtothisprojectpleasecontactbothsupervisors:
VincentGnanapragasam
[email protected]
Carola-BibianeSchoenlieb
[email protected]