Introduction

We report 10-year clinical outcomes of a prospective randomised controlled study on uni-compartmental knee arthroplasty using an active constraint robot.

Measuring the clinical impact of CAOS systems has generally been based around surrogate radiological measures with currently few long-term functional follow-up studies reported. We present 10 year clinical follow up results of robotic vs conventional surgery in UKA.

In computer assisted orthopaedic surgery, intraoperative registration is commonly performed by fitting features acquired from the exposed bone surface to a preoperative virtual model of the bone geometry. In cases where the acquired spatial measurements are unreliable or have been inappropriately chosen, the registration result can degenerate. Current performance indicators, such as the root mean squared (RMS) error and the spatial distribution of the registered feature errors may not be sufficient to warn the surgeon of such a case.

In this study, statistical analysis is applied to the registration outcomes of perturbed variants of a collected point set. In this way, it is possible to assess the ability of the original set to represent the underlying surface, taking into account the distribution of the points as well as errors introduced during the acquisition process. Confidence measures are calculated to predict the reliability of the original registration result and therefore the robustness of the point set itself.

For proof of concept, this method has been tested in simulation with a CT-generated tibia model. The algorithm was used to identify the 10 best performing of a population of 1000 randomly generated point sets. All registration outcomes produced by these point sets were found to be superior to those resulting from sets of the same size produced manually using an optimised point-acquisition protocol. Preliminary results suggest that this method, alongside the standard RMS and residual point error distribution, may be used to provide the surgeon with a reliable indication of registration outcome in the operating room.

A Prospective, randomised controlled trial demonstrates superior outcomes using an active constraint robot compared with conventional surgical technique in unicompartmental knee arthroplasty (UKA). Computer assistance should extinguish outliers in arthroplasty, with robotic systems being able to execute the preoperative plan with millimetre precision.

We used the Acrobot system to deliver tailor made surgery for each individual patient. A total of 27 patients (28 knees) awaiting unicompartmental knee arthroplasty were randomly assigned to have the operation performed either with the assistance of the Acrobot or conventionally. CT scans were obtained with coarse slices through hips and ankles and fine slices through the knee joint. Preoperative 3D plans were made and transferred to the Acrobot system in theatre, or printed out as a conventional surgical aid. Accurate co-registration was confirmed, prior to the surfaces of the femur and tibia being milled. The outcome parameters included measurements of the American Knee Society (AKS) score and Western Ontario and McMaster Universities Osteoarthritis (WOMAC) index. These measurements were performed pre-operatively and at six, 18 weeks, and 18 months post-operatively. After 18 months two UKA out of the conventional trial (n =15) had been revised into a total knee replacement (TKA), whereas there were no revisions in the Acrobot trial group (n = 13).

Using an active constrained robot to assist the surgeon was significantly more accurate than the conventional surgical technique. This study has shown a direct correlation between accuracy and improvement in knee scores at 6, 18 weeks and 18 months after surgery. At 18 months there continues to be a significant improvement in the knee scores with again a marked correlation between radiological accuracy and clinical outcome with higher accuracy leading to better function based on the WOMAC and American Knee Society Score.

Whilst computer assistance enables more accurate arthroplasty to be performed, demonstrating this is difficult. The superior results of CAOS systems have not been widely appreciated because accurate determination of the position of the implants is impossible with conventional radiographs for they give very little information outside their plane of view.

We report on the use of low dose (approximately a quarter of a conventional pelvic scan), low cost CT to robustly measure and demonstrate the efficacy of computer assisted hip resurfacing. In this study we demonstrate 3 methods of using 3D CT to measure the difference between the planned and achieved positions in both conventional and navigated hip resurfacing.

The initial part of this study was performed by imaging a standard radiological, tissue equivalent phantom pelvis. The 3D surface models extracted from the CT scan were co-registered with a further scan of the same phantom. Subsequently both the femoral and acetabular components were scanned encased in a large block of ice to simulate the equivalent Hounsfield value of human tissue. The CT images of the metal components were then co-registered with their digital images provided by the implant manufactures. The accuracy of the co-registration algorithm developed here was shown to be within 0.5mm.

This technique was subsequently used to evaluate the accuracy of component placement in our patients who were all pre-operatively CT scanned. Their surgery was digitally planned by first defining the anterior pelvic plane (APP), which is then used as the frame of reference to accurately position and size the wire frame models of the implant. This plan greatly aids the surgeon in both groups and in the computer assisted arm the Acrobot Wayfinder uses this pre-operative plan to guide the surgeon.

Following surgery all patients, in both groups were further CT scanned to evaluate the achieved accuracy. This post-operative CT scan is co-registered to the pre-operative CT based plan. The difference between the planned and achieved implant positions is accurately computed in all three planes, giving 3 angular and 3 translational numerical values for each component.

Further analysis of the CT generated results is used to measure the implant intersection volume between the pre-operatively planned and achieved positions. This gives a single numerical value of placement error for each component. These 3D CT datasets have also been used to quantify the volume of bone resected in both groups of patients comparing the simulated resection of the planned position of the implant to that measured on the post-operative CT.

This study uses 3D CT as a surrogate outcome measure to demonstrate the efficacy of CAOS systems.

Last year at CAOS UK we reported on the development of the Acrobot® Navigation System for accurate computer-assisted hip resurfacing surgery. This paper describes the findings of using the system in the clinical setting and includes the improvements that have been made to expedite the procedure. The aim of our system is to allow accurate planning of the surgery and precise placement of the prosthesis in accordance with the plan, with a zero intra-operative time penalty in comparison to the standard non-navigated technique.

The system uses a pre-operative CT-based plan to allow the surgeon to have full 3D knowledge of the patient’s anatomy and complete control over the sizes and positions of the components prior to surgery.

At present the navigation system is undergoing final clinical evaluation prior to a clinical study designed to demonstrate the accuracy of outcome compared with the conventional technique. Whilst full results are not yet available, this paper describes the techniques that are being used to evaluate accuracy by comparing pre-operative CT-based plans with post-operative CT scans, and gives initial results.

This approach provides a true measure of procedure outcome by measuring what was achieved against what was planned in 3D. The measure includes all the sources of error present within the procedure protocol, therefore these results represent the first time that the outcome of a navigated orthopaedic procedure has been measured accurately.

We used computer tomography (CT) to measure the outcome of knee-arthroplasty in our prospective double-blind randomised controlled study of our active constraint robotic system ACROBOT.

All patients in our trial had pre-operative CT scan and proprietary software used to plan the size, position and orientation of the implants. Post operatively a further CT scan was performed and measurement studies performed using 3 different methods of manipulating the CT dicom data.

Method 1, a quick and simple method of implant assessment that measures the varus-valgus orientation of the implants relative to the axes of the long bones

Two landmarks each are used to define the individual mechanical axis for both the femur and tibia, for consistency these landmarks are the very ones used in the planning stage on the pre-operative CT.

Landmarks are then placed on the implants in order to measure their tilt relative to the mechanical axes. An appropriate Hounsfield threshold (2800) was used to image the metal components. The angle between the individual mechanical axis and the prosthetic component was calculated.

Method 2, detailed and accurate comparisons between the planned and achieved component positions in 3D are made. Co-registration of the precisely planned CT based models with surface models from the post-op scan gives real measurements of implant position enabling the measurement of the accuracy of component in an all six degrees of freedom giving both translation and rotation errors in all three planes.

The process of alignment was achieved by surface-to-surface registration. An implementation of the iterative closest point algorithm was used to register matching surfaces on the objects to be registered. A polygon mesh of the implant, provided by the manufacturer, defined the surface shape of each size of implant. This was used both to define the planned position and to register to the post-operative scan. Method 3, in this study we quantified post-operative error in knee arthroplasty using one value for each component whilst retaining 3D perspective.

The position of the prosthetic components in the post-op scan is calculated and individual transformation matrix computed which is matched to the transformation matrices for the planned components.

The pre-operative CT based component positions were co-registered to the post-operative CT scan and values for the intersection (volumetric) between the digitised images (both planned and achieved) were calculated. Both the co-registered femoral and tibial component’s intersection was quantified with software packages supporting Boolean volume analysis

Method 1, the sum of the two, independently measured, angles allows an estimate of the post-operative alignment of the load bearing axes in the two bones.

Method 2, 3D CT allows precise measurements of the achieved position for each component in all three planes. Six values, three angular and three translational, define the achieved component position relative to the planned position.

Method 3, the greater the percentage intersection between the planned and achieved images, the greater the accuracy of the surgery. Owing to the shape of the components (large articular surface) large intersections demonstrate more accurate reconstruction of the joint line.

In the recent past the lack of a sufficiently accurate tool to plan and measure the accuracy of component placement has resulted in an inability to detect and study radiological and functional outliers and hence the hypnotised relationship between prosthetic joint placement and outcome has been difficult to prove.

CT offers us the ability to accurately describe the actual position and deviation from plan of component placement in knee arthroplasty. Whilst X-ray has the intrinsic problems of perspective distortion magnification errors and orientation uncertainties CT can be used to define ‘true’ planes for two dimensional (2D) measurements and permits the comparison in three dimensions (3D) between the planned and achieved component positions.

Hip resurfacing has advantages over hip replacement for younger, more active patients. However, it requires that surgeons learn new techniques for correctly cutting bone and positioning the components. Pre-operative planning systems exist for conventional hip replacement. Planning software for hip resurfacing is described, with the resulting plans available as a visual aid during surgery, or transferred to the Acrobot^® Navigation system for intra-operative guidance.

CT data is acquired from the top of the pelvis to immediately above the acetabulae in 4 mm slices, and from there down to just below the lesser trochanter in one mm slices. This keeps radiation doses low while providing high image quality in the important regions for planning. This is segmented semi-automatically, and bone surface models are generated.

Frames of reference are generated for the pelvis and femur, and the acetabular and femoral head positions are computed relative to these.

Prosthesis components are initially positioned and sized to match the computed anatomy. They can then be adjusted as required by the surgeon. While adjusting their positions, he is able to visualize their fit onto the bone to ensure good placement without problems such as femoral neck notching.

Twenty one hip resurfacings have been planned including two navigated cases. In addition, visualization of hip geometry for osteotomy and impingement debridement has been performed on 14 cases, giving the surgeon a good understanding of hip geometry prior to surgery. Initial evidence indicates surgeons find the planner useful, particularly when the anatomy is not straightforward.

Accurately planning the intervention and precisely measuring outcome in computer assisted orthopaedic surgery (CAOS) is essential for it permits robust analysis of the efficacy of these systems.

We demonstrate the use of low dose computer tomography (CT) radiation for both the planning and outcome measurement of robotic and conventionally performed knee arthroplasty.

Studies were initially performed on a human phantom pelvis and lower limb. The mAs (milliampere seconds) were varied from 120 to 75 at the pelvis and from 100 to 45 for both the knee and ankle whilst keeping the kV (kilovolt) between 120 and 140. Image quality was evaluated at the different doses.

The volumes scanned were defined on the scout film; they included the whole femoral head (0.5cm above and below the head), 20cm at the knee (10cm on either side of the joint line) and 5cm at the ankle (the distal tibia and the talus). Effective dose (mSv) was calculated using two commercially available software packages. This protocol was subsequently used to image patients in our prospective double-blind randomised controlled study of our active constraint robotic system ACRO-BOT.

With the reduction in the mA and scanned volume the effective dose was reduced to 0.761 mSv in females and 0.497 mSv in males whilst maintaining a sufficient image resolution for our purposes. We found that a mAs of 80 for the hip joint, 100 for the knee and 45 for the ankle was sufficient for imaging in both pre-op planning and pos-operative assessment in knee arthroplasty. This contributed on an average effective dose to the hip of 0.61 mSv, the knee 0.120 mSv and to the ankle 0.0046 mSv.

The results of our study show that we have considerably reduced the effective dose (0.8 mSv) to one third of the Perth Protocol (2.5mSv) by reducing the areas of the body scanned and adjusting the mA for the various parts of the body whist maintaining the x, y and z axis throughout the scan. The areas between the knee, hip and ankle that were not exposed to radiation are not strictly necessary for the planning of knee arthroplasty, but it is essential that the leg does not move during the scanning process. In order to prevent this leg was placed in a radiolucent splint. For post op three dimensional (3D) assessments only the knee component of the protocol is necessary.

This paper presents initial results of the Acrobot® Navigation System for Minimally Invasive (MI) Hip Resurfacing (HR) which addresses the problems of conventional HR. The system allows true MI HR – mini-mising the incision and tissue retraction required, and conservation of bone in contrast to other MI total hip procedures.

Pre-operative CT-based software allows the surgeon to plan the operation accurately. Use of CT gives the greatest accuracy, and is the only method which can give an accurate assessment of procedure outcome (planned versus achieved implant position). Intra-operatively, the bones are registered by touching points using a probe connected to a digitising arm. Next a series of tools is connected so that bone preparation and implant insertion is performed using on-screen guidance.

The accuracy of the registration probe is within 0.6mm, inside the acceptable margin for optical tracker systems. We have validated this acceptability using registration simulations leading to a protocol which restricts registration errors to within 1.5mm and three degree. These error margins are within those in the literature for acetabular component placement using optical tracker based systems (five degree inclination, six degree anteversion). No comparable data could be found regarding the accuracy of femoral component placement during computer-assisted HR.

The system is currently undergoing clinical tests at one alpha site, with three further beta sites planned for early 2006. The methods described by Henckel et al (CAOS International Proceedings 1994, pp. 281–282) are being used to evaluate the performance of the system, comparing pre-operative to post-operative CTs to obtain a true, accurate measure of performance.

The primary objective of this study was to evaluate the performance of the Acrobot^® Sculptor system in achieving a surgical plan for implantation of unicompartmental knee prostheses, compared with conventional surgery. The Acrobot^® Sculptor is a novel hands-on medical device, consisting of a high speed cutter mounted on a robotic device which the surgeon holds and directs.

A prospective, randomised, double-blind (patient and evaluator), controlled versus conventional surgery study was undertaken and has been fully reported in Journal of Bone and Joint Surgery (British), 88-B.

All (13 out of 13) of the Acrobot^® cases were implanted with tibio-femoral alignment in the coronal plane within ±2° of the planned position, while only 40% (six out of 15) of the conventionally performed cases achieved this level of accuracy.

There was also a significant enhancement in the extent of post-operative improvement, as measured by American Knee Society (AKS) Scores at six weeks, in the cases implanted with the Acrobot^®. The difference between type of surgery is statistically significant (p=0.004, Mann-Whitney U test). Operating time (skin to skin) is higher in Acrobot treated subjects, but the difference between the two types of surgery fails to reach significance.

The Acrobot^® System was found to significantly improve both accuracy and short term outcome in this investigation. By permitting the creation of bone surfaces that can be machined by means other than an oscillating saw, the Acrobot^® System paves the way for novel implant designs to be developed, facilitating bone conserving arthroplasty in the knee, hip and spine with a new generation of even less invasive but more reliable procedures.

The accuracy of prosthesis implantation is closely related to their function and longevity; we report the development of an active constraint robot for minimally invasive unicompartmental knee arthroplasty (UKA) using CT and knee scoring.

Method: Pre and postoperative CT scans are performed. Pre-op scan CT scans were used to plan the precise position of implants on the bones. The femoral and tibial bone cuts were then generated, together with the software boundaries that constrain the surgeon. This plan was then used to define the cutting planes of the ‘Acrobot’ active constraint device that we have developed.

The Postoperative CT scan was compared with the preoperative plan. The distance of the joint line from the hip and ankle joint, and its angulation and rotation were compared to the preoperative plan. In addition, the position of the implants relative to their planned position has been computed.

Results: No significant complications have been encountered. Using the postoperative CT scans, in no case is the implant more than 2mm or 2 degrees from the planned position.

Conclusions: The Acrobot system for UKA has completed its preliminary trial satisfactorily. It provides a hands-on operation but with robotic levels of accuracy, through a minimally invasive approach. By abolishing outliers, it improves outcomes in UKA replacement.

The Acrobot®, an active constraint “hands-on” robotic system, gives navigation cues to the surgeon, and also assists him in the surgery, using active software constraints if he tries to depart from the preoperative plan. It has just entered clinical trials. We report the first 5 cases.

The Acrobot® system for precision total knee arthroplasty comprises the following components:

1. A CT-based planning system

a gross positioning device with separate brakes and encoders, locked off for safety during the procedure,

4. The Acrobot’s software which:

imports the preoperative plan,

Each patient’s knee scores were monitored and postoperative CT scan was compared with the preoperative plan.

Seven robot assisted arthroplasties have been performed. No significant complications have been encountered. The Knee and Womac Scores show that the procedure is safe and comparable to conventional surgery in the early postoperative period. The envelope of error on postoperative CT scans has been within the accuracy of the method of measurement, at < 1 mm and < 10 without the outliers which haunt every clinical series.

The Acrobot® system for total knee arthroplasty has completed its preliminary trial satisfactorily. It provides a handson operation but with robotic levels of accuracy. It is suitable for conventional open surgery, but its real place will be in the arena of minimally invasive unicondylar knee arthroplasty, hip arthroplasty and resurfacing, and in the spine, where active constraint will prevent potentially dangerous surgical errors.