Abstract
Purpose
To characterize the knee kinematic profiles of total knee arthroplasty patient knees intraoperatively, before implant insertion, using principal component analysis.
Method
Ninety-two patientsreceived Stryker Triathlon total knee arthroplasty (TKA) implants. The Stryker surgical navigation system was used for all surgeries. The system was used to define rigid bodies representing the femur and tibia, and to track the three-dimensional movement of the knee joint during surgery. Each knee was moved through a passive range of knee flexion/extension before and after implantation of the arthroplasty components. The frontal plane (medial-lateral) movement of the knee joint through a range of 10 to 120 degrees of flexion before implantation was calculated for each knee using the joint coordinate system (referred to as the pre-implant knee kinematic curve). Visual inspection of these patterns indicated three predominant curve types: a backward S shape, a backward C shape and a valgus to varus shape. Each curve was subjectively categorized into one of these three categories. Principal component analysis (PCA), a multivariate statistical analysis technique, was applied to the pre-implant knee kinematic pattern data to objectively extract the major patterns of curve types within the 92 knees. Analysis of variance was used to compare the mean differences in PC scores between the curve shape groups to confirm visual categorization.
Results
Of the 92 patient curves, 13 had a backwards S shape, 14 had a backwards C shape and 20 had a valgus to varus shape. Forty of the knee were categorized as ‘other’. The first 3 principal components extracted from the knee kinematics curves cumulatively explained 99.7% of the variability in the original data, confirming 3 predominant curve types. The first PC captured an overall measure of varus/valgus throughout the flexion range. The second PC captured an S-shape pattern, and the third PC captured a C-shape pattern. Analysis of variance showed statistically significant differences between the four groups of knees for each PC (p < 0.0001, p <0.0001, p = 0.04 for PC1, 2 and 3 respectively), indicating significant differences between the groups based on the S and C shape patterns.
Conclusion
Visual inspection of pre-implant knee kinematic curves indicates that the knees of patients with severe arthritis have different patterns, the predominant patterns being S and C shape patterns. Principal component analysis was used to confirm these patterns quantitatively and to quantitatively show the differences between patient curve types. Principal component analysis is therefore a potentially powerful tool for a computerized characterization of the dynamic pattern of patient knee joints prior to total knee arthroplasty. Ongoing and future work will be used to link these curve types to outcome metrics, joint morphology and surgical technique. This will provide valuable subject-specific knee dynamic information for surgical planning and design.
