ProjectconceptsTheprojectconceptwasinitiallydevelopedforCTimageguidance,withthefollowingworkflow:1)ThepatientisinstalledontheCTscannertable,andimagesareacquiredforplanningpurpose.2)Therobotbaseisfixedtothepatientandthepatientcanbeprepared.3)Therobotisattachedtothebasepart,ideallyoutsidethescannertunnel.4) The patient is positioned again into the scanner to perform robot registration, robot automaticpositioningandthentheintervention,usingforcefeedbackteleoperation.Theradiologistcanchangetheneedleorientation,eithertoadjusttheneedlepathorcompensateforneedlebending.

Fig.1.Procedureworkflow.Fromlefttoright:stages1)to4).

Thisworkflowwas based on the use of a body-mounted system, to have natural physiologicalmotioncompensation at the surface of the patient. Such a system is sensitive to body attachment and gravityeffects.To limit thesedrawbacks,partial compensationof thesystemgravitycanbeconsideredusingasupporting system, inorder tomakea consensusbetween the slave robot supportingand thepatient’snaturalbreathing.Initialsolutions

Based on the previous workflow, we proposed several designs for the slave system. First, we madesystematicmechanical architecture synthesis to solve theproblem in all its generality, andprovide thesystem with 2T2R (2 translations, 2 rotations) mobility. This led to an original architecture (Fig. 2a),whichisacontributiontoneedlemanipulationdevelopment[MMT2014].Despiteaclearinterestforthistype of parallel architecture to perform the needle manipulation, several difficult shortcomings wereunderlined:thepresenceofisolatedsingularitieswithintherobotworkspaceforrotationsgreaterthan45deg.w.r.tthenormaltothepatient'sskin;thedifficultytointegratefouractuatorsandtheirtransmissionsgiven the required torque, even after optimization; the positioning of the actuators outside of the CTimagingplane.Itthenappearedthatthepositioningoftheneedleattheentrypoint,whichjustifiedthetwotranslationsinthe2T2Rsystem,couldpossiblybesimplified.Apossiblesolutionwastomakecoarsepositioningofthesystem base at the entry point, using table motions under the laser beam of the CT scanner. As aconsequence, the needle orientation was possible with 2R mobility only. This system was completelydesigned (Fig. 2c), including a needle manipulation device (NMD) for grasping, insertion, and forcemeasurement.ThesoftwarefortheCTtablecontrolwasalsoimplemented,incooperationwithSiemens.

Fig.2.Fromlefttoright,successivedesignsofthesystem:a)prototypeof2T2Rarchitecture;b)firstdesignbasedona2Rarchitecture;

c)finaldesignwithlasersintering/polymermaterials,includingsolutionsforneedlegraspingandinsertion,andforCTregistration.However, discussions with the IHU Scientific Board in 2015 led to conclude that: i) the wholesystemwouldbe toocomplex froman industrialperspective; ii) thewholesystemwouldbe tooexpensive.Then,wedecidednottofabricatethewholesystembutrathertofocusremainingtimeandfundsonthedesignandassessmentoftheNMD,whichistheoriginal(andpatented)partofthesystem.Needlemanipulationdevices,andlastyeardevelopments Fig. 2c is a CADviewof the last generationof 2R solution (July2015),which includes theNMDwhoseprinciple is basedon successive grasp and releaseof theneedle, combinedwithneedle insertionwhengrasped. The NMD [BIOROB2016] includes a cable actuated needle-grasping device, a force sensor[IEEEMech2016]and a translation mechanism actuated by piezoelectric motors. During last year, theremaining funds were entirely dedicated to its development, with the perspective that in thefuture this tool should be associated to differentmedical robots.Wespentmostof the remainingfundstohireanengineerforsixmonths,withtwoobjectives:1)tofabricateandimplementtheinitiallyproposedNMD;2) toproposeand implementanalternativetoolwiththesametechnologybutatmuchlowercost.ThesetwoNMDhavesimilarprinciple,basedonthreesubsystems:aneedleinsertiondevice,aforcesensor,andaneedle-graspingdevice.ThefirstNMD[Biorob2016]wasdesignedinordertomeetthefollowingrequirements:

• minimalsizeinordertobeutilizedwithinaCT-scannertunnel;• forcemeasurementalongtheneedle,upto10N(axial);• graspingofneedleswithdifferentdiameters,self-centering;• possibilitytograspandreleasetheneedle;• possibilitytomanuallyremovetheneedleatanymoment;• constructioncompatiblewiththeimagingmodalities,inparticularaslittlemetalaspossible.

Theresultingsystemwasimplementedusingadvancedrapidprototypingtechniques(Fig.3aand3b).Theinsertiondeviceincludesacustomforcesensorconstructedwithpolymermaterialsandstraingaugestoensuremedicalimagingtransparency[IEEETMech2016].Asthissystemwashighlyoptimizedbecauseofitsexpected integrationwithin thewholesystem, itsvolume isquite limited. Theactuator forgraspingwereplacedawayfromtheCTplane.ForneedleinsertionverysmallpiezoelectricactuatorswereselectedwithagoodcompatibilitywithCTimaging(Fig.3cand3d).

Fig.3.Fromlefttoright:a)needlegraspingdevicedetails;b)prototypeofneedlemanipulationdevice;c)CTimagesoftheprototype;

d)volumicrenderingofthetestbedobtainedfromastandardabdominalspiralacquisition.Inparallel,anewversionoftheNMDhasbeendeveloped.Thesystemisthoughttobeassociatedtothefluoroscopy Artis Zeego system. It is designed to serve as a tool mounted onto the end effector of amedical robot. ThisNMD is compatiblewith theKukaLBR IIWA that canbeused in an automatic or acollaborativeway.TherewasanothergoodreasonforthechoiceoftheKukarobot:Siemenshadsucha

robotat IHU,andused itwith theSiemensArtisZeego inother researchprojects.However, thoughweexpected to use it in the proteCT project, itwas finally not possible. Fortunately, in themeantime,weapplied and obtained an IDEX funding from Strasbourg University for two Kuka IIWA (they are beinginstalledwhilewewritethisreport).The new NMD includes the same force sensor and needle-grasping device, but its needle insertionmechanism isdifferent.The constraintson themetalparts and thebulkare less severewith theZeegothanwithaCT.Asaresult,thetransmissionmechanismandtheactuatorshavebeenchanged,andlowercostcomponentshavebeenselected.Inthisversion,low-levelcontrolisalsoimplementedusinglowcostsystems(anArduinoboardconnectedtoaRaspberrysystem).Asimplegraphicalinterfaceisavailableonatactiledisplay,inordertoallowtheelementarymotionsofneedlegrasping,releaseandinsertion.Thevaluesmeasured by the force sensor are also displayed to the user. This low cost control architecture,however,isnotcompatiblewithforcefeedbackteleoperation.Thissystemhasbeentestedinvivo(Fig.4bto4d)usingapassivesupportingarm.WehavetestedtheprincipleoftargetingwiththeSiemensiGuidesoftware (planning in the images, and targeting using the system laser and the Zeego alignment). Ofcourse, theuseofeither thecollaborativemanipulationmodeof theKukarobot,or itspositioningaftertheautomaticregistration,willmaketargetingmucheasier.

Fig.4.DesignanduseofthenewlowcostNMD,fromlefttoright:a)CADviewofthesystemwithsideandtopcoversremoved;b)toolinitsversioncompatiblewiththeKukaIIWA(heremountedonapassivearm)duringexperimentsintheArtisZeegoroom;c)experimentsonananesthetizedswine:insertionintotheliverusingSiemensiGuidesoftware;d)3Dreconstructedimagesofthe

needleinsertionexperiment,whichconsistedintargetingthetipofapre-insertedneedle.