Minimally invasive 3D data registration f in computer and robot assisted total knee arthroplasty 1 P. F. La Palombara M. Fadda S. Martelli M. Marcacci Labaratorio di Biomeccanica, Isthuti Ortopedici Rizzoli, via di Barbiano 1/10, 40136 Bologna, Italy Abstract~Computer and robot assisted surgery is concerned with the improvements achievable by using computer methods and robotic devices to plan and execute surgical interventions. The registration of different coordinate frames, often achie.ved through the matching of 3D data sets, represents a crucial step connecting planning and execution. Orthopaedic surgery already features a number of functioning applications which include registration routines relying on presurgicatty implanted fiducial markers. Replacing such invasive routines with non-fiducial registration procedures is regarded as a necessary step towards a minimisation of surgical invasiveness. A minimally invasive registration tech- nique based on the iterative closest point algorithm is presented and conceived for a specific computer and robot assisted orthopaedic reconstructive intervention, namely total knee arthroplasty. The whole surgical protocol is examined in detail and the experimental results, relative to tests performed on synthetic and animal specimens, are thoroughly reported and discussed. The authors indicate that the proposed registration approach is well-suited for the relevant application and appropriate for in vivo testing. Keywords---Registration, Iterative closest point, Computer assisted surgery, Total knee arthroplasty, Joint replacement Med. Biol. Eng. Comput., 1997, 35, 600-610 1 Introduction COMPb'TER ASSISTED surgery (CAS) is becoming a widely popular technolog'y and is now leaving laboratories and uni- versities and entmng surgical rooms. The application of computer techniques in the simulation, planning and execution of surgical interventions, aimed at increasing the rate of surgical success, is CAS's primary target. The advantages of CAS systems are widely recognised and orthopaedic surgery, in particular, can already feature a number of operative implementations. This is mostly due to a compelling need for high geometric accuracies and to some coexisting inherent characteristics of the skeletal system such as relative rigidity and fairly differentiated visualisation in standard diagnostic imaging technologies. Computer and robot assisted joint replacements (Ho et al., 1995; KAZA_~ZmES et al., 1995), osteotomies (CAPONET'rl and FANELH, 1993), screw and nail placements (LAvALLgE et al., 1994; NOLTE et al., 1994; 1995; VL~rr et al., 1995) and ligament recon- smactions (OR'n et aL, 1993) have been investigated in recent years. Our group has already contributed to the ongoing debate by submitting a computer integrated approach to total knee arthroplasty (TKA) (MARCACCI et al., 1995). The implant of knee prostheses, specifically TKA, is among the surgical procedures which would principally benefit from a computer integrated approach. A computer based planning system reduces the number of decisions the surgeon has to Correspondence should be addressed to Dr. La Pafombara. email: biomec@bo, nettuno.it First received 8 November 1996 and in final form 9 June 1997. 9 IFMBE: 1997 take intraoperatively by enabling them to be taken preopera- tively. As a consequence, it ensures a larger degree of safety in the surgical procedure and a reduction of the intraoperative time (RAND, 1993). Conversely, robotic assistance dmSng the execution phase is a means for improving the absolute accuracy in positioning and guiding surgical tools. The latter is a point of utter importance for implant fixation and long- term stability in TKA (PETERS and ROSENBERG, 1994). Successful implementation of CAS protocols requires a coherent integration of spatial data relative to a broad variety of imaging, sensing and actuating devices, each with its own coordinate system. This is why an accurate estimation of the geometric relationships between different coordinate frames and 3D data sets, normally referred to as 'registration,' plays a crucial role in virtually all CAS applications. In computer and robot assisted TKA (CRA-TKA) the registration step is a critical link between the planning and execution phases, because attainment of the same high geometric accuracy achieved during preoperative planning in the actual surgical execution is extremely important. Most of the current high accuracy protocols employ artifi- cial markers---fiducials----which need to be implanted before surgery. The fidueials are both identifiable in the preoperative images and accessible intrasurgicalty by means of a position transducer. Registration is then generally performed through fast and robust point-to-point matching algorithm.~ (TAYLOR et at, 1994; MITTELSTADT et aL, 1995; KIENZLE et al., 1995a; 1995b; LEA et al., 1995). Following a widespread trend towards a reduction in surgical invasiveness, intensive inves- tigation of non-fiducial registration techniques is currently in progress. Such effort concerns computer assisted orthopaedic surgery as well as other related fields (St,aON et aL, 1995a; 600 Medical & Biological Engineering & Computing November 1997