Description of Methodology
Thirteen fresh-frozen cadaver specimens were used for in vitro testing: one for the pilot experiment, and twelve for the subsequent measurements. The experimental setup was guided by the following sequence.
1. Mounting of the Reference Markers on the Tibia and the Femur
The specimen consisted of a complete limb, disarticulated at the level of the hip, and fresh-frozen. Commercially available optical reference markers (BrainLAB, Feldkirchen, Germany) were rigidly attached to the tibia, the femur, and the patella (Fig. 1). The reference markers carry spheres for reflecting infrared light. These markers serve a triple purpose: they can be located accurately with computed tomography both before and after implantation, they can be tracked by the three-dimensional motion-analysis system (Vicon Motion Systems, Los Angeles, California) for the purpose of following the motion of the rigid bodies, and they guide the surgical navigation system during insertion of the total knee prosthesis in the second part of the experiment.
2. Computed Tomography Scanning of the Frozen Specimen with Attached Markers
The specimens were transported, in the frozen state and in an isolating box, to the computed tomography scanner. We performed volumetric computed tomography scans on a sixty-four-row helical multidetector computed tomography (MDCT) scanner (General Electric Lightspeed VCT; General Electric, Milwaukee, Wisconsin). The images were obtained at 120 kV and 450 mA, with a slice thickness of 1.25 mm and a pitch of 0.5 mm per revolution. Raw data were processed with use of a bone filter. The images were stored in DICOM format onto a DVD. The computed tomography scans were analyzed with use of Mimics 11.02 medical image processing software and its MedCAD Module interface (Materialise, Haasrode, Leuven, Belgium) to create the surface reconstruction and identify the osseous landmarks (Fig. 2).
3. Thawing of the Specimen
The day before the experiment, the specimen was taken out of the refrigerator to allow sufficient time for thawing.
4. Initialization of the Surgical Navigation Software
At this stage, the surgical navigation software (BrainLAB) that was to be used for later implantation of the prosthesis was initialized, as the next step included amputation of the foot and resection of the proximal part of the femur and because important alignment landmarks for surgical references would be lost otherwise. The navigation software determined the center of the hip as the femur was moved through a number of circular motions, keeping the femoral head stabilized. In previous work in which biplanar radiographic measurements were compared with the computed center of the hip, a mean accuracy of 1.6 mm was shown19. The medial malleolus and the lateral malleolus were identified and located by the navigation software for later reference of the tibial mechanical axis.
5. Amputation of the Proximal Part of the Femur and the Ankle
The proximal part of the femur was amputated 32 cm proximal to the knee joint line. The foot was amputated 28 cm distal to the knee joint line. The bones were cleared of soft tissues and muscles over a distance of 12 cm.
6. Fixation of Femur and Tibia in Containers
The tibia and the femur were fixed in the holding containers of the mechanical knee rig with polymethylmethacrylate. The tibia was fixed in a position parallel to the container in the coronal and sagittal planes. The femur was positioned parallel to the container in the sagittal plane and under the physiologic valgus angle in the coronal plane. Screws were driven into the end of the bones to provide additional resistance strength during rotational movements.
7. Preparation of Tendon Fixation
On the medial side, the semitendinosus and semimembranosus tendons were sutured with use of number-5 Ticron sutures (Davis and Geck, Wayne, New Jersey) in preparation for the later attachment of constant load springs. On the lateral side, the biceps tendon was prepared in a similar way. Stronger fixation was needed for the quadriceps, however, because the pilot experiment revealed high forces; therefore, the tendon was cut 10 cm proximal to its attachment to the patella and then wrapped around a metal rod that had a sectional diameter of 5 mm. The tendon was sutured to itself with use of number-5 Ticron sutures (Davis and Geck) and reinforced with use of Mersilene tape.
8. Calibration of the Three-Dimensional Motion-Analysis Cameras
Five cameras (Vicon Motion Systems) were positioned on the medial side of the specimen to allow maximal visualization of the optical reference markers that were affixed to the tibia, femur, and patella. A special wand was used, calibrated to the distance between reflective spheres, thus enabling the software to calculate the relative distance of the cameras in space.
9. Mounting of Specimen in the Mechanical Knee Rig
The construct was then mounted on a dynamic knee simulator system that was based on the Oxford knee rig design but customized for this study (Fig. 3). This electromechanical system was designed to simulate and record the motions and loads in a knee joint during squatting. The hip joint can move up and down and can flex and extend. The ankle joint can move mediolaterally and has all three rotational degrees of freedom (flexion-extension, internal-external rotation, and abduction-adduction). Thus, this construction provides the knee joint with all six degrees of freedom. Only the flexion angle of the knee is controlled directly (by programming the hip position as a function of time). All other degrees of freedom are left free. Translations and rotations along these axes are governed by the geometry and anatomy of the joint. One actuator simulates the quadriceps muscle, and a second actuator produces vertical hip motion. The quadriceps actuator is positioned on the upper limb in a way that reproduces its anatomical location and thus also its moment arm with respect to the knee joint. Two constant force springs (50 N each) load the hamstrings on the lateral and medial sides of the tibia. They are fixed to the metal frame (representing the pelvis) to reproduce their biarticular function. Their position is such that their moment arms are similar to the in vivo situation. Sensors placed in line with the actuators detect the quadriceps force, the ankle forces and moments, and hip height relative to the ankle. A real-time data acquisition and closed feedback system (LabVIEW; National Instruments, Austin, Texas) was used to perform a squat with a certain hip velocity (given as a function of time) while simultaneously applying a quadriceps force on the knee to induce a vertical ankle force (defined as a function of time). The hip actuator is controlled by error feedback from the hip-position sensor with use of a proportional-integral-derivative controller. If the hip position is too high with respect to the programmed position at any instant during the flexion cycle (i.e., the hip is lagging behind), the actuator will be instructed to speed up, and vice versa. Likewise, the quadriceps actuator is controlled by error feedback from a six-axis ankle load cell under the ankle with use of a similar proportional-integral-derivative system. If the vertical ankle force is too low with respect to the programmed value at any instant during the squat, the actuator will be instructed to pull harder, and vice versa. Both control loops are nearly independent. Testing was performed at constant speed with ankle loads of 90 N, 130 N, and 180 N, from full extension to 120° of flexion. A process diagram of the rig is summarized in Figure 4.
10. Performance of the In Vitro Squat and Recording of Motion and Loads
The position of the rigid bodies consisting of the femur, tibia, and patella, with their respective reference frames, was followed as a function of time by the three-dimensional motion-analysis system. The loads on the ankle and on the quadriceps tendon were measured with calibrated load cells. Initial recording consisted of a positioning of the knee in full extension with the tibial container fixture loose along the vertical axis. The knee was then pulled into full extension by exerting a progressive load on the quadriceps. Static recording of the position of the rigid bodies was performed. A similar static measurement was performed with the knee positioned in 10° of flexion. The knee was then initialized at a position of 20° to 30° of flexion to start the dynamic measurements. A simulated squat was performed, and motion and loads were recorded during the full cycle.
11. Demounting of Specimen and Insertion of the Knee Prosthesis
A standard parapatellar medial incision was used to expose the knee joint. The patella was everted. The integrity of the cruciate ligaments and the articular cartilage was checked and recorded. A digital photograph of all specimens was made. Further steps for using the surgical navigation system were undertaken. Virtual surgical planning was executed, and osseous cuts were made accordingly. The knee prosthesis (Smith and Nephew, Memphis, Tennessee) was then inserted: a bicruciate stabilized design, with sacrifice of the ligaments, was used for the first six specimens; and a bicruciate-retaining design was used for the next six specimens. The components were fixed with polymethylmethacrylate, and the selection of insert thickness was performed as in surgical practice. The patella was not resurfaced. The arthrotomy was closed with a number-2 Vicryl suture (Ethicon, Johnson and Johnson, Somerville, New Jersey).
12. Mounting of the Specimen with Prosthesis on the Oxford Rig
The specimen with the prosthesis was mounted on the kinematic rig in an identical fashion to that of the specimen with the native knee.
13. Repetition of In Vitro Squat and Recording of Motion and Loads
All actions of step 10 were repeated.
14. Freezing and Computed Tomography Scanning of Specimen
The specimen was then frozen in extension, after the patella was cut loose and frozen separately (to avoid scattering from the prosthesis during the computed tomography scan). The specimen was taken in the isolating box in the frozen state to the computed tomography scanner. The scan was performed, following the identical protocol as previously described, with addition of the application of scatter-reduction software.
Description of Data Processing
The datasets of all specimens were handled in an identical sequence. The DICOM dataset of the native knee, produced by the computed tomography scan and taken before the in vitro experiment, was fed into three-dimensional-analysis software (Mimics; Materialise). All anatomical landmarks, ligament insertion sites, planes, and axes that were of interest were located on the computed tomography scan images (Figs. 2 and 5). A detailed description of all data points, planes, and axes, including interobserver and intraobserver variability, has been published previously20. The centers of the reflective optical markers were found by fitting a sphere to each marker. A Cartesian coordinate system was defined for each separate bone on the basis of surgically relevant axes (Fig. 5). As these Cartesian coordinate systems for the tibia, femur, and patella were defined, the coordinate system from the computed tomography scan was transformed into a separate coordinate system for each bone, with use of the surgically relevant planes and axes. This yielded a set of coordinates that allowed reconstruction of the joint line, analysis of quantitative anatomical relationships, and a comparison between the surface geometry of the native and the replaced knee. Also, the calculation of the coordinates of an identical preoperative and postoperative reference point allowed us to confirm that the reference frames had not moved during the experiment or during transportation. We followed the methodology of Grood and Suntay4 for the description of the kinematic behavior of the bones relative to each other; in addition, we created a joint coordinate system for all rotational descriptions. Medial femoral condylar translation was defined as the distance of the orthogonal projection of the center of the medial condyle in the horizontal plane of the tibia to the tibial transverse axis. The lateral femoral condylar translation was defined similarly (Figs. 6-A and 6-B). Below is a summary of the order of the mathematical processing of the data:Feed computed tomography-based data from the DVD into Mimics.Create surface model of the distal part of the femur, the proximal part of the tibia, and the patella.Locate all needed anatomical landmarks in Mimics software and export this list of coordinates, still in the computed tomography-based Cartesian coordinate system.Define Cartesian coordinate systems on the basis of the three optical tracking frames of the femur, the tibia, and the patella.Transform the coordinates of all anatomical landmarks from the computed tomography model into these three new Cartesian coordinate systems.Calculate the relative position of each anatomical landmark relative to the reference frame (femur, tibia, and patella) within the global Vicon coordinate system.Describe the position of each anatomical landmark as a function of time after filtering the Vicon data with a Woltering filter (Vicon) for mechanical (vibration and resonance), optical, or electronic noise.Define a single bone coordinate system for the femur, tibia, and patella (Fig. 5).For rotational kinematic description: Use the Grood and Suntay protocol for the joint coordinate system, the femoral mechanical axis and the femoral transverse axis as the body fixed axes, and the cross-product as the floating axis4.For translational kinematic descriptions, use projection planes as in definitions.
Validation
The system was validated through an accuracy and sensitivity analysis that included checks on reference-frame stability, repeatability, effect of ankle load, optical tracking accuracy, and component alignment.
1. Reference Frame Stability
During the tests, only the reflective markers affixed to the femur and the tibia were visible for the infrared cameras. The position of the bones was derived afterwards from the measured position of these reflective markers. As the experiment involved extensive handling and positioning of the specimens, frame stability was checked to confirm that the position of the markers relative to the bones had remained unchanged in the course of the experiment. A second computed tomography scan of the frozen specimens was performed after all experiments were finished. The relative position of selected reference points was compared with the initial position as measured on the first computed tomography scan. As the second scan was performed on the amputated specimen with the prosthesis in place, only a few points could serve as reliable landmarks. The center of the entry point in the cortex of one of the bone pins was selected as a primary reference point in both the femur and the tibia. The lateral epicondyle and the tip of the fibula were selected as secondary reference points in the femur and tibia-fibula unit, respectively. For each of the four control points, the distances between the point and the three reflective markers attached to the same bone were recorded on the first and the second computed tomography scans. A change in the distance between a control point and the markers would indicate that the frame with the reflective markers had moved with respect to the bone. The mean recorded changes in distance were smaller than 1 mm and thus of the same order of magnitude as the resolution of the computed tomography image. Only the lateral epicondyle showed an average position change of more than 1 mm (1.5 mm). This can be explained by the greater intraobserver variability with regard to this point20.
2. Repeatability
A basic analysis of the repeatability of the knee kinematics simulation was done on data obtained from cadaver specimens. The purpose of the analysis was to determine the effect of repeated squat trials on specimen kinematics when all test conditions were the same. Only native knee repeatability results are reported here because initial testing showed that native knees are more variable compared with replaced knees, which are generally more constrained. Only the test conditions with three or more squat trials were included for repeatability analysis. The repeatability of the parameters of tibial rotation, medial femoral condylar translation, lateral femoral condylar translation, tibiofemoral coronal alignment, ankle load, and quadriceps load were measured as functions of flexion angle. The repeatability of the system was quantified on the basis of previously described methods. For each set of repeated trials under the same test conditions, the mean kinematics curves were calculated. Then the residual errors between each trial curve and the mean curve were calculated, and the standard deviations of the errors were recorded as measures of variability. The worst-case, or largest, variability values were used to calculate coefficients of repeatability for system measurements. Assuming normality of the residual errors, the coefficient of repeatability was defined as two times the standard deviation of the residuals21. As such, the coefficient of repeatability values give approximate 95% confidence intervals for the kinematics parameters around their mean curve in each test condition. Four test conditions were analyzed, and their results are summarized in Table I. Each of the test conditions had a different specimen, target ankle load, and/or flexion range. On the basis of the results of the most variable conditions, 95% of the system measurements (from repeated squat trials in descent after 30° of flexion) would be expected to fall within the following ranges about a mean curve:Tibial rotation: 0.5°Medial femoral condylar translation: 0.8 mmLateral femoral condylar translation: 0.7 mmTibiofemoral coronal alignment: 0.6°Ankle load: 13 NQuadriceps load: 49 N
These ranges of variability are small. The results suggest that one squat trial at any particular condition gives a similar result as averaging three to four trials, within the ranges listed above.
3. Effect of Ankle Load
This analysis was performed on the knee kinematics simulator to find the effect of target ankle load on cadaver knee kinematics during a squat, relative to interspecimen variability. This information helped determine which ankle loads to study in the full experimental analysis.
Measurements versus flexion angle for specimens 2 and 3 were analyzed before and after total knee arthroplasty at 90 N, 130 N, and 180 N target ankle loads in descent. These loads were chosen as the extremes of possible test conditions. The variation among kinematics curves at different loads was recorded.
The measurements analyzed included tibial rotation, medial femoral condylar translation, and lateral femoral condylar translation, and their results are shown in Figure 7 here because they had the most variation. The other kinematics parameters showed less variation and are not discussed in this paper. Replaced knees also showed less variation than native knees and their results are also not discussed.
Figure 7 presents graphs of the results for native knees. The shapes of the curves at different loads were approximately the same within the same specimen, although the absolute values changed by small amounts. Tibial rotation and the two translations showed maximum differences between the 90 N and 180 N curves of less than 2° and less than 4 mm, respectively. Interspecimen differences exceeded the interload differences.
The effects of ankle load were small for tibial rotation, medial femoral condylar translation, lateral femoral condylar translation, and the other parameters not shown here, including those for replaced knees. The general patterns of the curves also were similar as ankle load changed. Therefore, any single ankle load between 90 N and 180 N should be able to demonstrate possible differences between different knees. Plots of the data showed that interspecimen variability was greater than interload variability in the conditions tested. The knee simulator itself also was noted to more often reach the target ankle load at loads of greater than 90 N. Also, the quadriceps load was less variable at ankle loads of less than 180 N. As a compromise for system stability, one physiologic target ankle load of 130 N could be chosen for study. This would reduce the number of squat trials and would also reduce the chance of damage to the specimens.
4. Optical Tracking Accuracy
Different Vicon system configurations were analyzed for their accuracy in tracking the displacements between two reflective markers. The configurations varied according to the number of cameras (four to eight), the layout of cameras (wide or narrow convergence angles), the camera lens focal length (6 to 12.5 mm) and target distance, target volume (approximately 50 to 120 cm3), the reflective marker diameter (9.5 to 14 mm), and marker displacement magnitude (0 to 20 cm). The displacements measured by the system were compared with the displacements measured with use of digital hand calipers that were accurate to 0.01 mm. Two markers were attached to the calipers, and the calipers were arbitrarily moved by hand in the motion-capture target area with the markers spaced at a locked distance. The cameras measured the distance over several seconds of video frames. The markers were then displaced by a known amount, as measured with use of the calipers, and the motion-capture system measured the new distances to find the displacements. Trajectories were not filtered in order to give worst-case data. The differences, or errors, between the displacements measured by the cameras and those measured with use of the calipers were quantified over each video frame for the unfiltered data. The worst-case raw data were then reanalyzed by first filtering the trajectories with the Woltering filter function (mean square error = 10) that was supplied with the Vicon Nexus software, as recommended by the manufacturer's instructions. The motion data were analyzed in the same way as described previously, and new error values were calculated.
The worst-case Vicon system configuration had a mean error and standard deviation of 0.06 ± 0.60 mm (n = 7000 frames) without filtering the marker trajectories. When the trajectories were filtered, the error was reduced, producing errors of 0.03 ± 0.19 mm (n = 7000). Filtering the data reduced the size and variability of error by approximately 50% and 70%, respectively. Prior to this study, we tested the Vicon system in expected worst-case laboratory conditions and found that the system can measure marker displacements within ± 0.60 mm for unfiltered trajectories and ± 0.19 mm for filtered trajectories. This assumes that all marker trajectories can be reconstructed throughout the motion-capture trial, meaning that the markers are not completely blocked from view. Measurement errors are expected to be smaller for slow movements and when there is minimal camera obstruction, such as in the camera configuration used for analyzing the knee kinematics simulator. Use of a Woltering filter (mean square error = 10) on raw trajectories was chosen for the Vicon data analysis of anatomical motions.
5. Component Alignment
All components were positioned with the help of a surgical navigation system (BrainLAB). The orientation in the coronal and sagittal planes was executed as planned on the virtual planning station, aiming at an orthogonal position to the femoral mechanical axis and the tibial mechanical axis in the coronal plane and an orthogonal position to the femoral mechanical axis for the femur and 3° of flexion to the tibial mechanical axis for the tibia, in the sagittal plane. The rotational alignment of the femoral component was based on the posterior condylar line, adding 3° of external rotation. The rotation of the tibial component was based on the femoral component, with the knee in full extension. A surface reconstruction of the prosthetic components, based on the second (postexperiment) computed tomography scan, was made, and the positions were calculated within the coordinate system of the femur and the tibia. These positions are shown in Table II. The mean actual positions of the tibial and femoral components in the coronal plane were within one degree of error from the desired position (0°). The maximum deviation was 3.2° on the tibia and 2.5° on the femur. The mean composite tibiofemoral coronal alignment position was 0.2°, as calculated from the postexperiment computed tomography scan. Theoretically, this composite tibiofemoral component position should match the limb alignment measured independently with the optical tracking frames. The last column of Table II shows the difference between the values obtained with these two independent methods. The reported deviations and differences between measurements reflect the following cumulative errors:Surgical navigation system errorSurgical block position errorSurgical bone-cutting execution errorMeasurement error on the postoperative computed tomography scanPositioning error on the knee rig (unwanted varus or valgus stress due to an incorrect cemented position of the femur or tibia in the containers)Ligament imbalanceMeasurement error with the Vicon systemCalculation error with coordinate transformations
As the reported deviations are small, it can be concluded that the methodological setup and sequence of measurements and calculations are robust. In addition, all specimens were positioned within 3° of varus and/or valgus error in the coronal plane, clinically considered as perfect alignment. In the sagittal plane, the mean flexion and standard deviation was 2.4° ± 3.2° for the femoral component and 1.4° ± 2.9° for the tibial component. In the horizontal plane (rotational alignment), the mean alignment for the femoral component was 0.3° ± 1.8°.
6. Study Weaknesses
The model has intrinsic weaknesses. Cadaver specimens are often hard to obtain. The specimens may be in a weakened state if the donor was elderly or had a chronic illness. Tissue quality can degrade over time. As the setup is labor-intensive and time-consuming, the number of tests that can be carried out per specimen is limited. The loaded simulated squat is a reproducible motion, but it certainly does not reflect the full spectrum of knee motions of daily life. Finally, the intricate interaction between hamstrings and quadriceps cannot be fully modeled, and the actual forces exerted in vivo by the hamstring muscles are unknown.
Statistical Analysis
Within each group of bicruciate-retaining and bicruciate stabilized specimens, the kinematics before and after total knee arthroplasty were compared with use of a general linear model analysis of variance for unequal sample sizes and also across flexion angles. Pairwise comparisons of data were made with use of the Tukey method. For all statistical tests, significance was set at p = 0.05.
Source of Funding
This experiment was carried out in the European Centre for Knee Research, sponsored by Smith and Nephew.
Plots of anteroposterior medial and lateral femoral translations (medial femoral condylar translation, lateral femoral condylar translation) versus flexion angle are shown in Figures 8-A and 8-B for bicruciate-retaining prostheses and in Figures 9-A and 9-B for bicruciate stabilized prostheses.
Native Compared with Bicruciate-Retaining Knees
Analysis of variance showed that, on the average, the tibial rotation of the replaced knee was 3.4° more internally rotated (p < 0.001) than it was in the native knee, while the medial femoral condylar translation was 3.0 mm more anterior (p < 0.001) than it was in the native knee. Quadriceps load was 48 N larger for the replaced knee (p = 0.037). Lateral femoral condylar translation was not significantly different between the native and replaced knees. No pairwise differences between native and replaced knees at individual flexion angles were detected. A significant relationship between flexion angle and lateral femoral condylar translation was found among native and replaced knees (p = 0.002), but not for medial femoral condylar translation. From 0° to 120° of flexion, in the loaded setting, the mean ranges of motion for tibial rotation, medial femoral condylar translation, and lateral femoral condylar translation were, respectively, 4.3°, 5.0 mm, and 6.9 mm for the native knee; and 4.8°, 4.3 mm, and 9.3 mm for the bicruciate-retaining replaced knee.
Native Compared with Bicruciate Stabilized Knees
Analysis of variance showed that, on the average, the tibial rotation of the replaced knee was 3.6° more internally rotated (p < 0.001) than it was in the native knee, while the medial femoral condylar translation was 2.5 mm more anterior (p < 0.001) than it was in the native knee. No significant difference in quadriceps load was found between the native and the replaced knees. No pairwise differences between native and replaced knees at individual flexion angles were detected. Medial femoral condylar translation and lateral femoral condylar translation were both found to be related to flexion angle in the native and replaced knees (p < 0.001). From 0° to 120° of flexion, the mean ranges of motion for tibial rotation, medial femoral condylar translation, and lateral femoral condylar translation were, respectively, all significantly higher for the replaced knee (p < 0.001) at 4.8°, 6.9 mm, and 6.8 mm for the native knee and 9.7°, 12.9 mm, and 16.3 mm for the bicruciate stabilized replaced knee.
Both bicruciate-retaining and bicruciate stabilized replaced knees produced loaded kinematics curves with shapes similar to the native knees. However, they also both showed some differences in the absolute kinematics values compared with the native knees, on the order of 3° to 4° more of tibial internal rotation and 2 to 3 mm more anterior translation of the medial femoral condyle. Overall, the kinematics of the bicruciate-retaining knees showed smaller deviations from those of the native knees. The bicruciate stabilized knee, in contrast, forced more femoral rollback and posterior translation than did the native knee, when both were tested in the loaded condition. The mechanical substitution with a cam and post of the intricate cruciate ligament system does not allow the knee to follow the changes that are induced in the native knee by muscle-loading.
In general, the kinematics of the replaced knee were not more variable than those of the native knee. In the case of bicruciate-retaining tibial rotation, the kinematics were even less variable than they were in the native knee.
Regarding quadriceps loading, the bicruciate-retaining knees showed a 48 N higher average load during the simulated squat than did the native knees, whereas the bicruciate stabilized knees did not show any significant difference. This may be explained by the different quadriceps moment arms in the two implants. The bicruciate stabilized knee showed more femoral rollback, which would lengthen the distance between the tibial insertions of the quadriceps and the tibiofemoral contact points. Effectively, this would give the quadriceps a larger moment arm with which to pull the knee into extension. Thus, for the same motions, a smaller quadriceps force would be needed in a bicruciate stabilized knee as compared with the force that would be needed in a bicruciate-retaining knee.
In conclusion, the model described allows accurate measurement of the position of the femur, tibia, and patella as a function of time, and, as such, it can be used to analyze passive and loaded knee kinematics, alignment, and ligament (an)isometry in the native knee and the prosthetic knee. Case-to-case comparison between the preoperative and postoperative setting is possible. System stability, repeatability, load sensitivity, and accuracy have been determined through repeated testing and independent double checks.
The study of knee kinematics through cadaver work contributes to a better understanding of the normal function and pathology of the joint and complements the results obtained through other in vitro and in vivo methodologies. 