NOViSE – a Virtual Natural Orifice Transluminal Endoscopic ......By inserting a flexible endoscope via a natural orifice such as the oesophagus, vagina, or anus (Fig. 1), and then

Accepted for publication in the International Journal of Computer Assisted Radiology and Surgery on 23/03/16

NOViSE–aVirtualNaturalOrificeTransluminalEndoscopicSurgerySimulatorPrzemyslawKorzeniowski,AlastairBarrow,MikaelHSodergren*,NielsHaldandFernandoBelloCentreforEngagementandSimulationScience,DepartmentofSurgeryandCancer,ImperialCollegeLondon,UK*DepartmentofSurgeryandCancer,ImperialCollegeLondon,UK

PrzemyslawKorzeniowski,ChelseaandWestminsterHospital,369FulhamRoad,SW109NH,[email protected]

Purpose:Natural Orifice Transluminal Endoscopic Surgery (NOTES) is a novel technique inminimallyinvasive surgery whereby a flexible endoscope is inserted via a natural orifice to gain access to theabdominalcavity,leavingnoexternalscars.Thisinnovativeuseofflexibleendoscopycreatesmanynewchallengesandisassociatedwithasteeplearningcurveforclinicians.Methods:WedevelopedNOViSE-the first force-feedback enabled virtual reality simulator for NOTES training supporting a flexibleendoscope.Thehapticdeviceiscustombuiltandthebehaviourofthevirtualflexibleendoscopeisbasedonanestablishedtheoreticalframework–theCosseratTheoryofElasticRods.Results:WepresenttheapplicationofNOViSEtothesimulationofahybridtrans-gastriccholecystectomyprocedure.Preliminaryresultsofface,contentandconstructvalidationhavepreviouslyshownthatNOViSEdeliverstherequiredlevel of realism for training of endoscopic manipulation skills specific to NOTES Conclusions: VRsimulation of NOTES procedures can contribute to surgical training and improve the educationalexperiencewithoutputtingpatientsatrisk,raisingethicalissuesorrequiringexpensiveanimalorcadaverfacilities. In the context of an experimental technique, NOViSE could potentially facilitate NOTESdevelopmentandcontributetoitswiderusebykeepingpractitionersuptodatewiththisnovelsurgicaltechnique. NOViSE is a first prototype and the initial results indicate that it provides promisingfoundationsforfurtherdevelopment.

Keywords—CosseratRod,Flexibleendoscope,NaturalOrificeSurgery,NOTES,VirtualRealitySimulator


2

IntroductionNOTESOver the last 30 years, laparoscopic surgery has become the standard approach formany operativeprocedures.Inordertopushminimallyinvasivetechniquesfurtheralongthespectrumtowardstrulynon-invasivesurgery,surgeonshavestartedusingflexibleendoscopyinprocedurestraditionallyreservedforrigidinstruments.Byinsertingaflexibleendoscopeviaanaturalorificesuchastheoesophagus,vagina,oranus(Fig.1),andthennavigatingtheendoscopethroughaninternalincisionintherelevantorgan,surgeonscangainaccesstotheabdominalcavityandareableto,forexample,removethegallbladder(cholecystectomy) or the appendix (appendectomy) leaving no external scars (incision/scar-lessprocedure). This emerging technique is known as Natural Orifice Transluminal Endoscopic Surgery(NOTES).Sinceiteliminatesexternalpostoperativewounds,itisarguedthatNOTESmayfurtherreduceoperationtrauma,recoverytime,clinicalcostsandimproveoverallcosmeticresults,therebypushingtheboundariesofminimallyinvasivesurgery(MIS)asweknowit[1-4].Aswithanynew,potentiallydisruptivesurgicaltechnique,thebenefitsofNOTESarestill tobefullyrealised,andthereremainsconsiderabledissentastoitstruebenefitsandrisks[5,6].TheNaturalOrificeSurgeryConsortiumforAssessmentandResearch(NOSCAR)establishedalistofpotentialbarrierswhichneedtobesurpassedbeforeNOTEScanbeincorporatedintoroutinepractice[7].Oneofthekeyissuesidentifiedbyspecialistswasthelackofefficient training programs available for clinicians and the extremely steep learning curve of NOTESprocedures.

Fig.1Trans-gastric(left),trans-vaginal(middle)andtrans-rectal(right)NOTESapproachesThenoveluseofaflexibleendoscopeinNOTESproceduresdifferssubstantiallyfrombothconventionalendoscopyandlaparoscopy.Inadditiontoenteringtheabdomenthroughanaturalorifice,theNOTEStechniquerequiresthesurgeontooperatetheendoscopeandanyassociatedinstrumentationthroughasingleaccesspoint,ratherthanthethreeorfourportscommontoalaparoscopicoperation.Althoughmany NOTES procedures are currently being performed in a hybrid fashion (i.e. with some trans-abdominalassistance),thereisasignificantlossofretractionandthein-lineinstrumentapproachthroughthe instrumentports in theendoscopehandpiece isunfamiliar tomost surgeons.Another significantdifferenceisthelackofagastrointestinallumentosupporttheendoscope.Thus,thedistalendoftheendoscope ismanipulated in theopenabdominalcavityusingthe incision (viscerotomy)site, internalorgans and gravitational force to navigate and position the instrument. The middle section of theendoscopeshaftcanunpredictablyrollandloopinsidetheabdomen.Asaresultofthesedifferencesinendoscopebehaviour,theapproachtoregionsof interestcanalsobeverydifferent incomparisontotraditionalendoscopictechniques.Takingalloftheaboveintoconsideration,itisclearthatperforming


3

NOTESproceduresrequiresanewsetofskills.Learningandpracticingtheseskillsdemandsanewsetoftrainingtaskssupportedbysuitablesimulatormodels.ThemostrecentsurveyoneducationandtraininginNOTES[8]reviews11non-animalstudies,8animalstudiesand6educationalprogramsforNOTES.Severalofthemdemonstrateconstructvalidity.Mostnotableisthe"ELITE"simulator-anex-vivo,full-scalereplicaofafemaleadultwithvarioustransluminalaccesspoints[9,10].ThesurveyalsostatesthatminimalworkhasbeencarriedoutinthefieldofVirtualReality(VR).

VRSimulationMedicalVRsimulatorsprovideasafeenvironmentinwhichclinicianscanrepetitivelypracticewithoutputtingpatientsatrisk.Theyhavebeenexpectedtobecomeanimportantpartofsurgicaltrainingsincetheearly1990s[11].Recentreviewsshowthat,althoughVRsimulationisnowsuccessfullyusedinvarioussurgicalspecialities, there isstillenormouspotential for furtherdevelopment [12-14].Arecentneedsanalysis foraNOTESVRsimulatorshowsthat there is indeed interest insuchtechnology [15].Whilstthere are well-established, validated commercial VR simulators available for flexible endoscopyprocedures(GIMentor-www.simbionix.com;EndoVR-www.caehealthcare.com)suchasLower/UpperGI,EndoscopicRetrogradeCholangiopancreatography(ERCP),EndoscopicUltrasound(EUS)orFlexibleSigmoidoscopy (FS), simulatingNOTESprocedures requiresmoreadvanceddynamicmodellingof thevirtual endoscope so that itmay operate in open abdominal cavities, and interact with surroundinganatomyinadifferentmanner.

In[16],Ahnandcolleaguesreporttheon-goingworkontheirVirtualTransluminalEndoscopicSurgeryTrainer–VTESTTM.TheirsimulatoraimstorecreateahybridNOTESprocedureusingarigidscopeandatrans-vaginalapproach. In[17],thesamegroupdescribestheirworkonaprototypehapticdeviceforflexible endoscope, but this is yet to be integrated into their VTESTTM system and no results of itsperformance are presented. Therefore, to the best of our knowledge, there are currently no force-feedbackenabled,eithercommercialorexperimental,VRsimulatorsforNOTESproceduressupportingaflexibleendoscope.

InthispaperweprovideadetailedtechnicaldescriptionofourNaturalOrificeVirtualSurgery(NOViSE)simulator, including its overall design, haptic interface, virtual flexible endoscope model, tool-tissueinteractions,aswellas implementationdetails.Next,wepresentresultsof theunderlyingendoscopemodelbehaviouralongwith itscomputationalperformance.This isfollowedbytheapplicationoftheNOViSEsimulatortoahybridtrans-gastriccholecystectomyprocedure(usingaflexibleendoscope)andasummaryofpreviouslyreportedinitialvalidationresults.

Ourmaincontributionisinthedevelopment(softwareandhardware)andintegrationofaforce-feedbackenabledvirtualrealitysimulatorforNOTEStrainingsupportingaflexibleendoscope,togetherwiththeadaptation of the underlying Cosserat rod mathematical model to make it inextensible andincompressible likea realendoscope,aswellasvariousperformance-optimisation features toensurereliableandstablereal-timeoperation.

MethodsNOViSEisafirstprototypewhosemainfocusisonteachingtheendoscopicmanipulationskillsrequiredfor NOTES. The crucial components of such a simulator are the custom built haptic device and theunderlyingmathematicalmodelof thevirtual flexibleendoscope.As there iscurrentlynoestablishedcurriculum for NOTES, NOViSE was developed in close collaboration with experts who previouslyconductedalargenumberofNOTESanimalandboxmodeloperations[18,8],aswellasaseriesofhumanNOTEStrials.

Basedontheirexperience,theprocedurestepswhicharethemostdemandingintermsofendoscopemanipulationwerecarefully identified.Theseare,namely:navigation into theabdomen;clippingandcuttingofacysticductandartery;anddissectionofthegallbladderfromtheliverbed.Theexpertsalso


4

identifiedthe followingNOTES-specificsetofskillsandchallenges,whichweretobethefocusof thecurrentNOViSEprototype,togetherwithasetofrelevantperformancemetrics(seerelevantsubsectionbelow):

• Openspacemanipulationsintheabdominalcavity• Navigationwithalackofgastrointestinallumen• Singleaccesspoint• In-lineinstrumentapproach,i.e.instrumentsalignedwiththemainendoscopeshaft• Limitedtissueretraction

HavingidentifiedthefocusofthefirstNOViSEprototype,sometrade-offswererecognisedinordertostreamline the development process and improve computational performance,without affecting theintendedtrainingoutcomes,namely:

• Static,non-deformablemodelsofoesophagus,stomachandintestines• Static,pre-deformedmodelofaretractedliver• Gallbladderandconnectivetissuemodelledasmass-springsystems• Empiricaltuningofthe(bio)-mechanicalparametersoftheendoscopeandgallbladder

Thefollowingsectionsprovidefurtherdetailsoftheimplementationandjustificationfortheabovetrade-offs,includingasummaryofpreviouslyreportedpreliminaryvalidationstudieswheretheircorrectnessandrelevancewasalsoassessed[19].

SimulatorSet-upNOViSE (Fig. 2) consists of a real-time software simulation and a physical, force feedback human-computer-interface (hapticdevice).Thesoftware iswritten in Java,withperformancecriticalsectionsimplementedinC/C++.Itcanefficientlyrun,exceedinghapticinteractiveratesonamodernmid-rangePCorlaptop.Thesimulatordisplayisdividedintotwoparts.Ontheright,theusercanseetheendoscopiccameraview.Onthe left there isanexternal,optional"aerialview",whichcanbeturnedon/offandfreelymanipulated.

Fig.2ThecompleteNOViSEset-up(left).Theoverviewofthesimulationsoftwaresub-systems(right).


5

HapticDeviceWhilst thereexistavarietyofhaptic interfaces forendoscopicsimulation [20],weundertookanovelredesignwiththeaimofportabilityandaffordability.Ourhapticdevice(Fig.3left)comprisesanenclosedblack box of dimensions approximately 55x26x18cm, into which passes a hose (1.5m long, 15mmdiameter).Thehosecanbepushedorpulledthroughtheopening(totaltravel22cm)androtatedfreely.Inside theenclosure, theendof thehose isdirectlycoupled toa15:1planetarygearboxandaservomotordeliveringacombinedtotaltorqueof+/-2.55Nm.Thismotorismountedonalowfrictionlinearraildrivenbyanidenticalmotorconnectedviaatensionedtootheddrivebeltanda24mmpulleyresultingin+/-14N linear forceoutput.Both linear (14N)and rotational (2.55Nm) force feedback significantlyexceedtherequirementsforendoscopicNOTESprocedures.Theseweremeasuredby[17]andpeakat4.77Nforlinearandonly0.03Nmforrotationalfeedback.

At the proximal end of the hose, a 3D printed plastic replica of a standard endoscopic handpiece isattached(Fig.3right).Itconsistsoftwoforce-feedbackenabledthumbwheels,twoopticallytrackedthinwiresrepresentingtheendoscopictoolwiresandtwopushbuttons.Additionally,adoublefootpedalisplaced on the floor and can be used to activate endoscopic instruments, e.g. diathermy. ThethumbwheelsareactuatedusingacommonBowdencablearrangementbytwoservomotorslocatedinthemainenclosure.

Fig.3Thehapticdeviceconnectedtothedataacquisitiondevice(left).Aclose-uptothehandpiece(right).

VirtualFlexibleEndoscopeTheshaftofthevirtualendoscopecanbepushed,pulledandrotatedthroughmanipulatingthehapticdevice.Thetipofthevirtualendoscopeissteerable,withitsbendcontrolledusingtwothumb-wheelsonthehandpiece.Thevirtualscopeisequippedwithalightsource,acameraandtwoworkingportsthroughwhichdifferentinstruments(actuators)maybeinserted.Currently,operatorscanchoosefromfourtypesofvirtualactuators:grasper,clipper,scissorsanddiathermytool.Theirinsertion/removaliscontrolledbytwophysicalwiresinsertedinthetwoportsofthehandpiece(Fig.3,right).

Avitalpartofoursimulatoristheunderlyingmathematicalmodeloftheone-dimensionaldeformablebody (elastic rod) responsible for the behaviour of the virtual flexible endoscope. Elastic rods arecharacterized by having large non-linear deformations even if the local strains are small. Thischaracteristic,aswellasconsiderationofmaterialtwistandthefactthatmanyrodspracticallydonotstretch,makestheirdynamicsimulationchallenging,especiallyinreal-time.Physically-basedinteractiveapproachestoelasticrodsrangefrommass-springmodels[21-23],rigidmulti-bodyserialchains[24,25],splinebasedformulations[26],toCosserattheorybasedmodels[27-32].Forourflexibleendoscope,we


6

chosetheCoRdEmodelbySpillmannandTeschner[29].First,becauseitisbasedonCosserattheory–asolidtheoreticalfoundationconsideredasafinalstepintheformulationofamoderntheoryofelasticrods [33]. Second, because CoRdE is a fast, dynamic and elegant solutionwith an explicit centrelinerepresentation,which facilitates the simulationof contacts. Theexplicit centreline also simplifies theoverall implementation, internal frictioncalculationsandrodvisualization[29,30].CoRdEbehaviour isindependent of the underlying discretization. Additionally, thismodel was previously applied by ourgroupto thesimulationofcathetersandguidewires incardiovascular interventionswithgoodresults[34].

InCoRdE,aquaterniongoverningthematerialframes𝐑𝒋 ∈ ℝ𝟑𝐱𝟑(cross-sectionorientation)isplacedinbetween the neighbouring mass-points 𝐫𝒊 ∈ ℝ𝟑 consisting of the centreline as shown in Fig. 4. Bycomparingthematerialframes,thebendandtwistdeformationscanbequantifiedandthetotalelasticenergyoftherod𝑉canbederived.Byminimizing𝑉,thebendingandtwistingstrainratesarecoupledtogetherinaunifiedmanner.Thetwistdeformationisbalancedoutbythebenddeformationandviceversa,which results in the looping phenomenon.We refer the reader to the original publication forfurtherimplementationdetails[29].

Fig.4MaterialframesadaptedtotheCosseratrodcentrelineIntheoriginalmodel,thepenaltymethodwasusedtogovernthestretchofthecentreline.Duetothis,a reduction in the rod’s stretch and compressibilitymay introduce additional stiffness to the systemrequiringasmallertime-step.Toaddressthis,wemodifiedtheoriginalmodeltomakeitinextensibleandincompressible like a real endoscope. The penalty method was replaced by a chain of distanceconstraints. For accuracy andperformance, the constraints Jacobianswere arranged in a tri-diagonalbandedsystemofequations.Asthissystemcontainsonlyequalityconstraints,itisefficientlysolvedusingGaussianeliminationwithpartialpivotinginaglobalmanner(i.e.alldistanceconstraintsatonce)asafirststepofeachsolveriterationbyaconventional linearalgebralibrary(LAPACK,www.netlib.org). Inthenextstep,thecollisionconstraintsincludingCoulombianfrictionapproximationareappliedlocally.Ifneeded,thewholeprocessmayberepeatedmultipletimestoimprovesimulationaccuracy.

OurmodifiedCoRdEimplementationenablesreal-timesimulationathapticinteractiverates,providingan efficient, unified bending, twisting and collisions handling, aswell as a very fast response to usermanipulationsandaneasyparameterizationofthemechanicalpropertiesoftherod.Itallowsforthephysicalpropertiesofourvirtualendoscope,suchasmass,diameter,resistancetobendingandtwistingtobederivedempirically,underthesupervisionofexpertclinicians, tomatchthebehaviourofarealflexiblescope.

Tool-TissueInteractionsDuringthedesignstage,incollaborationwiththeNOTESexperts,aseriesoftestswereconductedatoidentify which organs and/or anatomical structures are relevant to the training of endoscopemanipulation skills. In a hybrid transgastric cholecystectomy, the endoscope is introduced via the


7

oesophagus,ittravelsinsidethestomachandfinallygoesthroughtheviscerotomysite(incisiononthewall of the stomach) to the abdominal cavity. Once in the abdominal cavity, the endoscope and itsactuatorscancollideandinteractwiththegallbladder,liver,fattyconnectivetissuebetweenthesetwo,aswellasthesmallandlargeintestineandthepancreas.

Having identified the relevant anatomy, theexperts compared two versionsof the first phaseof theprocedurewithandwithoutdeformationofoesophagusandstomach.Duetothehollownatureoftheseorgans,theirdeformationsweresimulatedusingmass-springmodels.Theexpertsagreedthat,althoughthedeformationsaddedtothevisualplausibilityintheauxiliaryview,theywerehardlynoticeableintheendoscopicviewalone,andthereforeweredeemedirrelevanttothetrainingofmanipulationskillsforNOTES.

Afterpassingthroughthestomachandreachingtheabdominalcavity,arigidlaparoscopeiscommonlyusedtoretracttheliverandthegallbladder.Thisinvolvesdeformationoftheliver,connectivetissueandgallbladder. Instead of attempting to model liver deformation in real-time, which was deemedunnecessarygiventhefocusofthecurrentprototype,we“pre-modelled”theretractedliverusinga3Dpackageasshownin(Fig.5,right).

Regardingthegallbladder,a3Dpolygonalmodelwasdiscretizedintoatetrahedralmeshconsistingof1194 mass-points, 3615 tetrahedrons and 5914 connecting springs using TetGen(www.tetgen.berlios.de). A mass-spring model (MSM) [35,36] was implemented to simulate itsdeformationasitinteractswiththeendoscopeandtheactuators.Itsbodycanbeprobed,grasped,anditscysticarteryandductcanbeclippedandcut.Additionally,thegallbladdercanalsoberetractedinaconventionalwayusingarigid laparoscope(hybridNOTESprocedure). Itsphysicalproperties,suchasspringstiffnessanddampening,werevisuallyandhapticallytunedtoapproximatethebehaviouroftherealanatomybasedonthejudgementofseniorcliniciansinteractingwiththevirtualanatomythroughthehapticdevice,whilstachievingreal-timeperformance.Althoughlessaccuratethancontinuumbasedmethods suchas FEM,MSMcanhandle largedeformationsaswell as topologicalmodificationswithrelativeease.Inspiteofmodelbehaviourbeingdependentontheunderlyingdiscretizationandspringstiffnessanddampingparametersnotdirectlycorrelatingtomeasurablereal-worldvalues,MSMmodelsarestilloftenusedtosimulatedeformableorgansastheyarefastandeasytoimplement.

TheconnectivetissuebetweenthegallbladderandtheliverisalsoapproximatedbyaMSMthatcanbedissected using the diathermy tool or scissors. Its physical propertieswere also fine-tuned by seniorclinicianswhilst interactingwith the virtual anatomy through the haptic device. The deformations ofotherabdominalorganssuchasthe intestinesandpancreaswereuniformly judgedbytheexpertsasratherirrelevantfromthepointofviewofmanipulationsduringtransgastriccholecystectomy.

Collisiondetectionbetweentheendoscopeoritsactuatorsandsurroundingtissuesisbasedonadynamicbounding volume hierarchy (BVH) of axis-aligned bounding boxes (AABB) as suggested in [37]. Forperformancereasons,onlytissueswhichareintheregionofinterestoftheoperator,i.e.reachablebytheendoscope(Fig.5,right),areconsideredduringcollisiondetection.Gravityasaconstantforceinthesupine position has been incorporated to both the gallbladder and the endoscope by consideringaccelerationduetogravity(g=9.81m/s2)inthecorrespondingdirection.


8

Fig.5Acutthroughatetrahedralmeshofthegallbladder(left).TheabdomentissuesreachablebytheendoscopewrappedinanAABBBVHtree(right).

Multi-threadedImplementationThesimulatorimplementationtargetedmulti-coreCPUs.Itfollowsataskparallelismapproach,inwhichthesimulationtasks(rodphysicscomputations,collisiondetection,deformablebodiessimulation,scene-graphtransformations)aredistributedacrossdifferentthreads.Differenttaskscanrunatdifferentratesrelativelytotheendoscopephysicstask.Thecollisiondetectiontaskrunsinparallel,onetoone,astepbehind,butthedeformablebodytaskrunsapproximatelyat1/3rateoftheinstrumentphysics.Weuseaformofdoublebufferingtoreducethesynchronizationoverheadandpreventconcurrencyissuessuchasvisualflickeringorphysicsinstabilities.Ourexperimentsshowthattaskparallelism,inthiscase,resultsinapproximately50%betterperformancecomparedtodataparallelism,whichspreadsmass-pointsandquaternionsacrossthethreads.Thisiscausedbyasmallersynchronizationoverhead,fewerCPUcontextswitchesandbettercacheutilization.

WealsocarriedouttestsusingaGPUimplementation(IntelCore22.66GHz,NVidiaGeForceGTX560).Theendoscopesimulationposesarathersmallprobleminthecontextofmassivelyparallelcomputationsstandardsandnoperformancegainswerenoted.Intermsofsoft-bodysimulation,areasonablex4-x5speed-upwasachieved for thegallbladder simulation.However,whenconsideringcollisiondetectionandinteractionsbetweentheorgansandthetools,therawcomputationalpowerofaGPUwaslargelymitigated by the overhead caused by memory transactions, kernel launches and CPU/GPUsynchronization.Hence,itwasdecidedtoinsteaddevelopNOViSEtargetingmodernCPUs.

MetricsTheNOViSEsimulationsoftwarecomputesandcollectsaseriesofperformancemetrics(Table1)whichwere derived during the analysis by the expert clinical collaborators of the particular tasks of hybridtransgastric cholecystectomy and endoscopic manipulation skills for NOTES. Once the procedure iscomplete,thesimulatorcanpresentandprocessthemetricstogenerateacomma-separated(CSV)filewithanadequatelayoutsuitabletobeimportedintoMSExcelorSPSSforfurther,moredetailedanalysis.Inaddition,foreachcompletedtask,anassociatedbinaryfilecontainingtherecordedmotionsofthehapticdeviceandofthevirtualendoscopeisgenerated.Fromthese,severalsupplementarymetricscanbeextractedorstoredperformancesplayedbackforfurtherstudy.

Results Inthissectionwepresentbriefqualitativeandquantitativeresultsofourelasticrodimplementationintermsof its behaviour andperformance.Next,wedescribe the application of the abovemethods inNOViSEtosimulateahybridtransgastriccholecystectomyprocedure.Finally,wesummarisetheresultsofaninitialfaceandcontentvaliditystudy.


9

Table1MetricscollectedduringthesimulationForalltasks:

• Taskcompletiontime• Pathlengthofthetiptraversedduringthetask• Maximumandaverageforcesappliedtotissuesbyendoscope• Maximumandaveragevelocitiesandaccelerationsoftheshaft• Maximumandaveragehapticforce-feedback

Forclippingandcutting(sub)-tasks:• Clipping/cuttingdistancefromtheindicatedpoint• Clipping/cuttinganglebetweenthetoolandthesurface• Numberofclippings/cuttings• Degreeofinstrumentprotrusionduringtheoperation

Forgallbladderdissectiontask:• Numberofdiathermyactivations• Totaltimediathermyactivationtime• Timediathermyactivatedontarget/non-targettissues• Percentageoftimeburningnon-targettissue

RodInextensibilityandIncompressibilityInordertodemonstratetheimportanceofinextensibilityandincompressibility,arodconsistingof256(e.g.255mmlong)pointswasgraduallyinsertedintoarandomrigidanatomymodelinawaytoaggravateits compression.Ourmodified rod compressedby less than0.5%. In contrast, theCoRdEmodelwithhighest possible (but stable) stretching Young’s modulus 𝐸,, compressed by nearly 6%. Such acompressionisnotonlyclearlynoticeable,asevidentinFig.6,butalsoresultsinaverydifferentbendandtwistdeformation.Thecomputationaltimerequiredbyourincompressiblerod(0.86msperiteration,includingcollisions)wasonly6%longerthanthatoftheoriginalCoRdEmodel(0.81ms).

Fig. 6Rod compression testby insertion intoa virtual vascularmodel.At theTop: theoriginalCoRdEmodel.AttheBottom:ourinextensibleandincompressiblemodification.Thepercentageofcompressionisgiveninbrackets.Thegreenspheresrepresentcollidingmass-pointsandtheredones-non-colliding

ComputationalperformanceTheentirecomputationalperformanceduringtheprocedurewastestedonanAsusN55slaptop(Win7x64, IntelCore i72.2GHz,8GBRAM,NVidiaGeForceGT555M).The totalcomputational timeof thevirtual endoscope (Cosserat forces, constraints, integration) consistingof 100mass-pointswasbelow


10

0.2ms.Thecollisiondetectionrunsinsynconaseparatethreadandslightlyslowsdowntherodphysics(0.23ms).Themass-springmodelforthedeformablegallbladderwastheslowestpartofthesimulationrequiringnearly0.75msperupdate.Thedisplaywasupdatedatastandard60framespersecond.InFig.7,we present the average computational times for each particular simulation (sub)-task: endoscopephysics,collisiondetectionanddeformablebodysimulation.

Fig.7ComputationaltimesofdifferentcomponentsonAsusN55sLaptopDue todifferenthardwareplatformsand test casesused, the computationalperformanceofour rodcannotbedirectlycomparedtootherCosseratrodimplementationsfoundintheliterature.However,approximate comparisons indicate that the performance of our approach is similar to the originalstretchableCoRdE[29].Yet,consideringtheinextensibility,ourrodwasroughlyx4,x38,x24timefasterthantheapproachespresentedin[30],[38]and[39],respectively.However,thesemodelshaveotheradvantages,forinstance,theeliminationofthepenaltymethodinparallelconstraints,improvedstabilityorbettercontacthandling.Table2summarizestheseresults.

Table2Comparisonofcomputationaltimesinrespecttoothermodels

Model PCCPUTimestatedinpaper(ms/pts)

Time(ms/100pts) Speed-up

Ourinextensiblemodel Core22.66GHz 0.147/100 0.147 x1.00

OriginalCoRdE[29] Xeon3.80GHz 0.131/100 0.131 x0.89

InextensibleCoRdE[38] Core23.00GHz 2.26/40 5.65* x38.4*

DiscreteElasticRods[30] Core22.66GHz 0.34/75-0.42/67 0.45*-0.67* x3.06-x4.56*

PositionBasedElasticRods[39] N/A 1.06/30 3.53* x24.0*

*Approximatedtimes

HybridTransgastricCholecystectomyWe chose to simulate a hybrid trans-gastric cholecystectomy since laparoscopic cholecystectomy(gallbladderremoval)isoneofthemostprevalentsurgical interventions.ItwasalsoamongstthefirstNOTESproceduresand it iscurrentlyoneway inwhichNOTES isperformed inclinicalpractice. It isahybrid operation as laparoscopic assistance for visualization and retraction is coupledwith a flexibleendoscope.Thesimulationstartswiththeendoscopepartiallyinsertedintotheoesophagus.Itisdividedintothreemaintasks:navigationviathestomachtotheabdomen(Fig.8),clippingandcuttingofthe

EndoscopePhysicsTask

CollisionDetectionTask

DeformableBodyTask

Mass-spring 0.74

Collisiondetection 0.23

Integration 0.018

Constraints 0.049

Cosseratforces 0.098

00.10.20.30.40.50.60.70.8

Time(m

s)

Timesofsimulationtasks(ms)


11

Calot'striangle(Fig.9)andgallbladderdissectionusingthediathermytoolandthegrasper(Fig.10).Theoperatorisguidedbyglowingmarkersindicatinganoptimalpath,anincision(viscerotomy)site,clippingpoints/anglesandconnectivetissue.

During the first task (Fig.8), theoperatorneeds to find theviscerotomysite locatedonasideof thestomachandnavigatethescopethroughitintotheabdominalcavity.S/heisnotrequiredtopiercethestomachastheincisionisalreadypresentandrepresentedbyaglowingredring.However,navigatingthroughtheringisnottrivialasitrequiresacombinationofbimanualmotionsoftheendoscopeshaftandhandpiececontrols.

Fig.8Task1-Navigationfromthestomachintotheabdomenviatheincision(viscerotomy)site(glowingredring)Afterenteringtheabdomen,theoperatorcanproceedtothesecondtask-clippingandcuttingofthecysticarteryandduct(Fig.9).TheoperatorneedstolocateananatomicalregioncalledCalot'striangleandstartbyclippingthecysticarteryfirst.Theoptimalclippingpointisindicatedbyabluemarkerwheretheoperatormustinsertaclippingtool,positioningitsjawsasclosetothebluemarkeronthearteryaspossible,whilemaintaininga rightanglebetween the jawsand theartery.Next, s/heneeds toplaceanothercliponthearteryandcutbetweentheclipsusingscissors.Thisclipandcutprocessisrepeatedonthecysticduct.Thekeytocompletingthistaskefficientlyistocorrectlynavigateandpositionthetoolrightfromthestartsothatalltheclippingandcuttingcanbedonewithouthavingtomanipulatetheendoscope.Thisway,allthesixpointsofinterestshouldbewithinreachbyonlyadjustingthetipofthescopeandinserting/removingtheactuators.


12

Fig.9Task2-Clippingthecysticartery

Havingclippedandcutthecysticductandtheartery,theoperatorcanprogresstodissecttheconnectivetissuebetweenthegallbladderandtheliverbedusingthediathermytool(Fig.10).Theconnectivetissueisrepresentedbyredglowinglinesegmentsthatareburntbytheoperatoractivatingthediathermytoolclosetothem.Activationofthediathermyneedstobepreciseandaccurateinordertoburnaslittleofothernon-targettissuesaspossible.

Fig.10Task3-Gallbladderdissectionusingdiathermy.Theredlinesegmentsrepresenttheconnectivetissue.Anauxiliarylaparoscoperetractingthegallbladderisvisible.Asmentionedabove,thesimulatedprocedureishybrid,whichmeansthatthereisstillonelaparoscopicinstrumentdeployed in theconventionalwayused to retract thegallbladder inorder togetabetterexposureoftheconnectivetissue.Thisretractioniscontrolledusingthekeyboardbytheassistantgiventhedirect command topreventbias.After removing the connective tissue, theoperator canuse the


13

actuatorgraspertoholdthegallbladderandpullitoutthroughthestomach(Fig.11).Atthispointtheprocedureiscompleted.Theoperatorisnotrequiredtoclosetheviscerotomysite.

Fig.11Completingtheprocedurebypullingoutthegallbladderusingthegrasper.In between certain tasks, the screen occasionally fades out and the simulation is paused.When thishappens, the operator is asked to adjust the insertion of the hose so that s/he will have enoughinsertion/retractionavailabletocompleteeachtaskwithoutreachingthelimitofthehapticdevice.

ValidationAninitialverificationoftheclinicalrealismandaccuracyoftheinstrumentswaspreviouslycarriedoutbyobtainingsubjectivefeedback(faceandcontentvalidity)throughaquestionnairefrom14surgeonsindifferentspecialities[19].FourofthemwerequalifiedasNOTESexpertswhohaveperformed10ormorehuman or animal-model NOTES procedures independently. A summary of the validation results ispresentedbelow.

NOViSEshowedgoodoverallfacevalidity(Fig.12).Inthequestionnaire,63%ofresponsestostatementsregardingtherealismofthevirtualendoscopeand67%ofresponsestostatementsregardingthevisualrealismoftheendoscopiccamerawere‘agree’or‘stronglyagree’.Theparticipantsweremostcriticaloftheoveralllookandfeelofthehardware.64%ofparticipantsstated(i.e.agreedorstronglyagreed)thatthesimulator,aswellasthedifficultyofthesimulatedprocedurewererealistic.


14

Fig.12Facevalidityresults–participants’responses[19]NOViSE showed also good content validity (Fig. 13). In the questionnaire, 74% of responses to thestatementsassessingtheusefulnessandrangeoftheindividualtasksfortrainingwere‘agree’or‘stronglyagree’.Majority(86%)oftheparticipantsstated(i.e.agreedorstronglyagreed)thatNOViSEisausefultrainingtoolforNOTESand71%thattheywouldrecommendittoothers.

Fig.13Contentvalidityresults–participants’responses[19]Regarding construct validity, expertswere faster and had better economy ofmovement (i.e. used ashorterendoscopicpathlength)thannovicesinallbutthefirsttask.Duetothenon-normaldistributionofdata,anon-parametricMann-WhitneyUtestwasusedtocomparetheperformancebetweengroups.Thestatisticalsignificancewassetatp<0.05anditwasdemonstratedforthefollowingtasksandmetrics:timeandeconomyofmovement(EoM)fromexitingstomachtoapplicationoffirstclip(74svs194sp=0.04,50cmvs191cm,p=0.01),timeandEoMfromapplicationoffirstcliptostartofdissection(83svs228sp=0.04,17cmvs134cm,p=0.04),timefromapplicationof lastcliptocompleteddissectionofgallbladderfromliverbed(333svs683sp=0.02,250cmvs527cm,p=0.04).Theremainingmetricsdidnotdemonstratestatisticallysignificantdifferencesbetweenexpertsandnovices.

0%

20%

40%

60%

80%

100%

Endoscopebehaviour

Visualization Hardwarelookandfeel

Difficultyofprocedure

Overallrealism Allresponses

StronglyDisagree Disagree Neutral Agree StronglyAgree

0%

20%

40%

60%

80%

100%

IndividualtasksandtheirrangeareusefulinNOTEStraining

OverallthesimulatorisausefultrainingtoolforNOTES

Iwouldrecommend thesimulatortoothers

Allresponses

StronglyDisagree Disagree Neutral Agree StronglyAgree


15

Fig.14Constructvalidityresults–taskcompletiontimes(left)andeconomyofmovement(right)[19]

LimitationsandFutureworkA main limitation of NOViSE is that it currently simulates only transgastric hybrid cholecystectomy.Supportingawider rangeofprocedures (e.g. appendectomy) andapproaches (trans-vaginal, possiblytrans-rectal),aswelladvancedendoscopic interventionsisanimportantnext-stepin itsdevelopment.GiventhatNOViSEisafirstprototype,itexhibitssomehardwareandsoftwarelimitations.Replacingthephysicalendoscopehoseforalighterandsofteroneisrequired,togetherwithmodificationstothedesignofthehandpiece.

During the virtual procedure, there is lackof fat tissue surrounding theCalot’s triangle. The stepsofcreationandclosureoftheviscerotomysiteareomitted.AddingthesefeaturesandtaskscouldrecreateamorecomprehensiveNOTESexperience.Thesoft-bodymodelusedforsimulatingthegallbladder isrelatively simple,whilst all theotherorgansare static anddonotdeform. Itwouldbe interesting toinvestigatefurtherifusingmoreadvanceddeformationmodels(suchasFEM),notjustforthegallbladder,butpossiblyalsofortheoesophagus,stomach,andliverwouldmakeaperceptibledifferencetotheuserexperienceinthecontextofthewholeprocedure.Manipulatingthepatient’spositionontheoperatingtable,asitisdoneinflexibleendoscopy,andadjustingtheresultinggravitationalforce,couldbeanotherpotentialfurtherrefinementofthesimulator.

Althoughinitialtestsinthecurrentprototypeshowednobenefitsfromusingamassively-parallelGPUimplementation,consideringtherapidprogressinthefieldandthepossibleneedforfurtherandmoreadvancedorgandefamationmodelling, thismayneed tobe re-evaluated.Morespecifically,aunifiedphysicsapproach[40]mightbeworthexploring.

Additional validation studies are foreseen after implementing the suggested modifications to bothhardwareandsoftwarecomponentsofNOViSE.Wewouldliketoevaluatethefeasibilityofourplatformfor the simulationof advancedendoscopicprocedures suchas EndoscopicMucosalResection (EMR),EndoscopicSubmucosalDissection(ESD)andPerOralEndoscopicMyotomy(POEM).

ConclusionsInthispaperwehavepresentedaprototypevirtualrealityforce-feedbackenabledsimulatorforNOTES,whichsupportsatrans-gastrichybridcholecystectomyprocedureusingaflexibleendoscope.Atthisstageofdevelopment,thefocushasbeenonteachingofspecificendoscopicmanipulationsrequiredforNOTES,yetNOViSEhasestablishedpromisingfoundationsforfurtherdevelopment.Theoperatorinteractswith

Experts Novices

Navigation 44 68

1stClipping 74 194

Clipping/Cutting 83 228

Dissection 333 683

0200400600800

100012001400

time(s)

Experts Novices

Navigation 44 78

1stClipping 50 191

Clipping/Cutting 17 134

Dissection 250 527

0100200300400500600700800900

1000

pathleng

th(cm)


16

thevirtualendoscopeviaacustombuilthapticdevice.Thebehaviourofthevirtualflexibleendoscopeisbased on the Cosserat theory, allowing for realistic recreation of bending and twisting of the virtualendoscope,aswellasguaranteeingafastresponsetousermanipulationsandincorporatinggravityasaconstant force in the supine position. The efficient, multi-threaded implementation enables thesimulationtorunefficientlyonanoff-the-shelfPCorlaptopathapticinteractiverates.

Inapreliminaryvalidationstudy [19],NOViSEhas showngoodoverall faceandcontentvalidity,withimprovementssuggestedtothefeelofthehapticdeviceanddesignofthehandpiece.Participantsagreedthat NOViSE is sufficiently realistic, it can be a useful training tool for NOTES and that they wouldrecommendittoothers.NOViSEalsodemonstratedearlysignsofconstructvalidity.Expertswerefasterandhadbettereconomyofmovementsthannovicesin3outof4tasksthannovices.

TheseinitialresultsindicatethatNOViSEcanrecreateatrans-gastrichybridcholecystectomyprocedure,potentially contributing to surgical training and to improving the educational experience for NOTES,withoutputtingpatientsatrisk,raisingethicalissuesorrequiringexpensiveanimalorcadavericfacilities.Moreover,consideringthatNOTESisstillanexperimentaltechniquewithoutanestablishedcurriculum,NOViSEoffersthepossibilitytofacilitate itsdevelopmentthroughVRsimulation.For instance, inpre-operativeplanningorprototypingof new surgical devices.NOViSE couldalsopotentially support thewideradoptionofNOTESandadvancedendoscopyproceduresbykeepingpractitionersuptodatewiththesenovelminimallyinvasivesurgerytechniques.

Acknowledgment:Theauthorswishtothankthestudyparticipants.

Funding:TheprojecthasbeenfundedbytheEPSRC(EngineeringandPhysicalSciencesResearchCouncil)InstitutionalSupportFund(GrantcodeEP/J018201/1).

ConflictofInterest:PKorzeniowski,ABarrow,MHSodergrenandNHalddeclarethatthereisnoconflictofinterest.DrBelloreportsgrantsfromEngineeringandPhysicalScienceResearchCouncil,duringtheconductofthestudy;heisalsoafoundershareholderandunpaidDirectorofConvincisLtd,outsidethesubmittedwork.

Informedconsent:Informedconsentwasobtainedfromallindividualparticipantsincludedinthestudy.

References1. Spivak H, Hunter JG (1997) Endoluminal surgery. Surg Endosc 11 (4):321-325 2. MacFadyen BV, Jr., Cuschieri A (2005) Endoluminal surgery. Surg Endosc 19 (1):1-3. 3. Richards WO, Rattner DW (2005) Endoluminal and transluminal surgery: no longer if, but when. Surg Endosc 19 (4):461-463. 4. Ponsky JL (2006) Endoluminal surgery: past, present and future. Surg Endosc 20 Suppl 2:S500-502. 5. Rattner DW, Hawes R, Schwaitzberg S, Kochman M, Swanstrom L (2011) The Second SAGES/ASGE White Paper on natural orifice transluminal endoscopic surgery: 5 years of progress. Surg Endosc 25 (8):2441-2448. 6. Moris DN, Bramis KJ, Mantonakis EI, Papalampros EL, Petrou AS, Papalampros AE (2012) Surgery via natural orifices in human beings: yesterday, today, tomorrow. American journal of surgery 204 (1):93-102. 7. Rattner D (2006) Introduction to NOTES White Paper. Surg Endosc 20 (2):185. 8. Moghul MR, Sodergren MH, Clark J, Teare J, Yang GZ, Darzi A (2013) Education and Training in NOTES: A Systematic Review. Surg Innov 20 (3):282-291. 9. Gillen S, Wilhelm D, Meining A, Fiolka A, Doundoulakis E, Schneider A, von Delius S, Friess H, Feussner H (2009) The "ELITE" model: construct validation of a new training system for natural orifice transluminal endoscopic surgery (NOTES). Endoscopy 41 (5):395-399. 10. Gillen S, Fiolka A, Kranzfelder M, Wolf P, Feith M, Schneider A, Meining A, Friess H, Feussner H (2011) Training of a standardized natural orifice transluminal endoscopic surgery cholecystectomy using an ex vivo training unit. Endoscopy 43 (10):876-881. 11. Satava RM (1993) Virtual reality surgical simulator. The first steps. Surg Endosc 7 (3):203-205


17

12. Gallagher AG, Traynor O (2008) Simulation in surgery: opportunity or threat? Irish Journal of Medical Sciences:177:283–287 13. de Visser H, Watson MO, Salvado O, Passenger JD (2011) Progress in virtual reality simulators for surgical training and certification. The Medical journal of Australia 194 (4):S38-40 14. de Montbrun SL, Macrae H (2012) Simulation in surgical education. Clinics in colon and rectal surgery 25 (3):156-165. 15. Sankaranarayanan G, Matthes K, Nemani A, Ahn W, Kato M, Jones DB, Schwaitzberg S, De S (2013) Needs analysis for developing a virtual-reality NOTES simulator. Surg Endosc 27 (5):1607-1616. 16. Ahn W, Dargar S, Halic T, Lee J, Li B, Pan J, Sankaranarayanan G, Roberts K, De S (2014) Development of a Virtual Reality Simulator for Natural Orifice Translumenal Endoscopic Surgery (NOTES) Cholecystectomy Procedure. Studies in health technology and informatics 196:1-5 17. Dargar S, Lam B, Horodyski C, Sankaranarayanan G, De S (2014) A Decoupled 2 DOF Force Feedback Mechanism for the Virtual Translumenal Endoscopic Surgical Trainer (VTEST. Studies in health technology and informatics 196:86-88 18. Clark J, Sodergren M, Noonan D, Darzi A, Yang GZ (2009) The natural orifice simulated surgical environment (NOSsE): exploring the challenges of NOTES without the animal model. Journal of laparoendoscopic & advanced surgical techniques Part A 19 (2):211-214. 19. Korzeniowski P, Brown DC, Sodergren MH, Barrow A, Bello F (2015) Validation of NOViSE – a novel Natural Orifice Virtual Surgery Simulator. (Under review). Surg Innov 20. Samur E, Flaction L, Bleuler H (2012) Design and Evaluation of a Novel Haptic Interface for Endoscopic Simulation. Ieee T Haptics 5 (4):301-311. 21. Rosenblum RE, Carlson WE, Tripp E (1991) Simulating the structure and dynamics of human hair: Modelling, rendering and animation. The Journal of Visualization and Computer Animation 2 (4):141-148. 22. Wang F, Burdet E, Dhanik A, Poston T, Teo CL (2005) Dynamic thread for real-time knot-tying. World Haptics Conference: First Joint Eurohaptics Conference and Symposium on Haptic Interfaces for Virutual Environment and Teleoperator Systems, Proceedings:507-508 23. Selle A, Lentine M, Fedkiw R (2008) A mass spring model for hair simulation. Acm T Graphic 27 (3). 24. Choe B, Choi MG, Ko HS (2005) Simulating complex hair with robust collision handling. Paper presented at the Proceedings of the 2005 ACM SIGGRAPH/Eurographics symposium on Computer animation, Los Angeles, California, 25. Hadap S (2006) Oriented strands: dynamics of stiff multi-body system. Paper presented at the Proceedings of the 2006 ACM SIGGRAPH/Eurographics symposium on Computer animation, Vienna, Austria, 26. Theetten A, Grisoni L, Andriot C, Barsky B (2008) Geometrically exact dynamic splines. Comput Aided Design 40 (1):35-48. 27. Bertails F, Audoly B, Cani MP, Querleux B, Leroy F, Leveque JL (2006) Super-helices for predicting the dynamics of natural hair. Acm T Graphic 25 (3):1180-1187. 28. Gregoire M, Schomer E (2007) Interactive simulation of one-dimensional flexible parts. Comput Aided Design 39 (8):694-707. 29. Spillmann J, Teschner M (2007) CORDE: Cosserat Rod Elements for the Dynamic Simulation of One-Dimensional Elastic Objects. Symposium on Computer Animation 2007: Acm Siggraph/ Eurographics Symposium Proceedings:63-72 30. Bergou M, Wardetzky M, Robinson S, Audoly B, Grinspun E (2008) Discrete elastic rods. Acm T Graphic 27 (3). 31. Casati R, Bertails-Descoubes F (2013) Super Space Clothoids. Acm T Graphic 32 (4) 32. Pai DK (2002) STRANDS: Interactive simulation of thin solids using cosserat models. Comput Graph Forum 21 (3):347-352. 33. Goss VGA (2003) Snap buckling, writhing and loop formation in twisted rods. University College London, 34. Korzeniowski P, Martinez-Martinez F, Hald N, Bello F (2014) Simulation of Catheters and Guidewires for Cardiovascular Interventions Using an Inextensible Cosserat Rod. Biomedical Simulation 8789:112-121 35. Meier U, Lopez O, Monserrat C, Juan MC, Alcaniz M (2005) Real-time deformable models for surgery simulation: a survey. Computer methods and programs in biomedicine 77 (3):183-197. 36. Nealen A, Muller M, Keiser R, Boxerman E, Carlson M (2006) Physically based deformable models in computer graphics. Comput Graph Forum 25 (4):809-836. 37. Bergen Gvd (1998) Efficient collision detection of complex deformable models using AABB trees. J Graph Tools 2 (4):1-13.


18

38. Spillmann J, Harders M (2010) Inextensible elastic rods with torsional friction based on Lagrange multipliers. Comput Animat Virt W 21 (6):561-572. 39. Umetani N, Schmidt R, Stam J (2014) Position-based elastic rods. Paper presented at the ACM SIGGRAPH 2014 Talks, Vancouver, Canada, 40. Macklin MM, M.; Chentanez, N.; Kim, TY.; (2014) Unified Particle Physics for Real-Time Applications. ACM Transactions on Graphics (TOG) 33 (4):104

Documents

NOViSE – a Virtual Natural Orifice Transluminal Endoscopic ......By inserting a flexible endoscope via a natural orifice such as the oesophagus, vagina, or anus (Fig. 1), and then