Background:
One of the primary goals of total knee replacement is to restore native knee function. Tibiofemoral kinematics provides an objective measure of knee function and have been used to characterize and compare knee function among native (i.e., healthy), replaced, osteoarthritic, and anterior cruciate ligament (ACL) deficient knees. Tibiofemoral kinematics specifically refers to the relative rigid body motions of the tibia with respect to the femur in all six degrees of freedom. Quantification of clinically meaningful tibiofemoral kinematics requires a joint coordinate system where motions are free from kinematic crosstalk errors.
One common method to determine tibiofemoral kinematics is to capture single-plane fluoroscopic images of a patient activity and determine the relative position and orientation of the tibia with respect to the femur. This method involves the use of 3D model-to-2D image registration in which projections of 3D models of the native knee (or replaced knee) are fitted to the silhouette of their respective components in the image. The most common software used to perform this registration is JointTrack, of which there are two different versions (JointTrack Manual and JointTrack Auto).
Because the precisions of the two different JointTrack programs are unknown, the first objective was to determine the overall precision of both programs in determining the AP positions of the femoral condyles for an example set of TKR components. The rationale behind using the AP Positions as the primary dependent variable is that it is one of the most common variables used to determine tibiofemoral kinematics in the literature and uses a method (lowest point) that has been validated by many research groups. Furthermore, the TKR knee was analyzed because JointTrack was specifically designed to perform image registration with the replaced knee.
While the AP positions of the femoral condyles provide information about tibiofemoral kinematics in primarily one degree of freedom (internal tibial rotation), there are still relative motions in four degrees of freedom other than flexion that need to be determined. As previously stated, quantification of clinically meaningful tibiofemoral kinematics requires a joint coordinate system (JCS) where motions are free from kinematic crosstalk errors. For the JCS to be free of kinematic crosstalk errors, two key requirements are that the body-fixed flexion-extension (F-E) and internal-external rotation (I-E) axes must coincide with the functional axes and that the femoral and tibial Cartesian coordinate system origins must lie on the functional axes. However, the International Society of Biomechanics (ISB) recommends the joint coordinate system (JCS) described by Grood and Suntay, where the axes are not functional and is therefore subject to kinematic crosstalk errors. However, these errors are unknown.
Therefore, the second objective of this study was to determine whether a JCS constructed using functional body-fixed axes (termed FUNC JCS) reduced kinematic crosstalk errors compared to the ISB JCS. As these JCSs were constructed for the native knee, this investigation was done in the native knee using JointTrack Manual per the information gathered from the first chapter of this thesis (Appendix A).
Methods:
For the first study on JointTrack precision, fluoroscopic images of 16 patients who performed a weighted deep knee bend following TKR were analyzed. JointTrack Manual and JointTrack Auto were used to perform 3D model-to-2D image registration and determine the absolute positions and orientations of the femoral and tibial components in each image. The AP positions of the femoral condyles were determined using the lowest point method. Precision was found by performing image registration 3 times for each patient and computing the variability in the AP positions measured for each trial. Intraclass correlation coefficients (ICCs) were also determined for both JointTrack programs. As the native knee requires the use of JointTrack Manual (due to degraded image and 3D model quality), further analysis was performed to inform the methods of the second study (Appendix A).
In the second study, fluoroscopic images of native knees in 13 subjects performing a deep knee bend were analyzed. JointTrack Manual was used to perform 3D model-to-2D image registration and determine the absolute positions and orientations of the femoral and tibial components in each image. Relative rigid body motions of the tibia with respect to the femur in all six degrees of freedom were calculated for both the ISB JCS and the FUNC JCS using a Cardan angle sequence and corresponding transformation matrix.
Results:
Overall precision for the JointTrack Manual program was 3 times worse than the JointTrack Auto program for both medial and lateral AP positions of the femoral condyle (0.97 mm and 0.91 mm versus 0.34 mm and 0.38 mm, respectively; p < 0.0001 for both). ICC values for the Auto program indicated good to excellent agreement (range: 0.82 – 0.98); whereas ICC values for the Manual program indicated only moderate to good agreement (range: 0.58 – 0.82). Precision using JointTrack Manual was significantly improved for determining the medial and lateral AP positions when performing three trials and taking the average of the three trials (0.58 and 0.70 mm, respectively; p = 0.0001 for medial and p = 0.0288 for lateral).
Results for the second study show that tibiofemoral kinematics using the FUNC JCS fell within the physiological range of motion in all five degrees of freedom excluding flexionextension. Internal rotation of the tibia with respect to the femur averaged 13° for the FUNC JCS versus 10° for the ISB JCS and motions in the other four degrees of freedom (collectively termed off-axis motions) were minimal as expected based on biomechanical constraints. In contrast, offaxis motions for the ISB JCS were significantly greater; maximum valgus rotation was 4°, maximum anterior translation was 9 mm, and maximum distraction translation was 25 mm, which is not physiologic.
Discussion:
It is not surprising that JointTrack Auto has better precision and reproducibility as indicated by higher interobserver ICCs than JointTrack Manual for measuring the tibiofemoral kinematics of the TKR knee. JointTrack Manual requires considerably more operator intervention and considerably more time to perform image registration than JointTrack Auto. Therefore, the use of Auto over Manual is strongly recommended. However, if JointTrack Manual must be used, which is necessary for 3D models developed from image segmentation, then it is recommended to perform Manual image registration 3 times and take the average of the 3 trials for better precision.
For the second study, the FUNC JCS achieved clinically meaningful kinematics by significantly reducing kinematic crosstalk errors compared to the ISB JCS and is the more suitable coordinate system. Moving forward, the ISB is well advised to update their recommendation, which was published 20 years ago, so that the updated recommendation reflects the latest knowledge.