Effects of radiograph projection parameter uncertainty on TKA kinematics from model-image registration

Model-image registration techniques have been used extensively for the measurement of joint kinematics in vivo. These techniques typically utilize an explicit measurement of X-ray projection parameters (principal distance, principal point), which is easily done for prospective studies. However, there is vast opportunity to derive useful information from previously collected clinical radiographic films where the projection parameters are unknown. The purpose of this study was to determine variation in measured knee arthroplasty kinematics when the X-ray projection parameters were unknown, but bounded. Based on the clinical radiographic protocol, a nominal principal point was chosen and eight additional points +/-2 and +/-5 cm in the horizontal and vertical directions were defined. Tibiofemoral kinematics were determined for all nine projection parameter sets for a series of 10 lateral radiographs. In addition, the principal distance was varied +/-15 cm and tibiofemoral kinematics were determined for these two projection sets. Measured joint kinematics varied less than 0.6 degrees and 0.4 mm for +/-2 cm variations in principal point location, and 0.7 degrees and 0.6 mm for +/-5 cm variations in principal point location. Measured joint kinematics varied less than 0.6 degrees and 0.7 mm for +/-15 cm variations in principal distance. Variation in X-ray principal point and principal distance over clinically bounded ranges has a small effect on knee arthroplasty kinematics computed from model-image registration with high-quality clinical radiographs.