Application of Subject-specific Models for Surgical Planning of Knee Implants

Total knee arthroplasty is a surgical procedure with high long-term reliability, yet up to a fifth of primary implant patients remains unsatisfied. This high patient dissatisfaction rate reveals that the current surgical procedure does not always result in the desired functional outcome. Total knee arthroplasty typically aims to restore an overall neutral leg alignment, such that the line going from the centre of the hip joint to the centre the ankle joint passes through the knee centre. Currently, this neutral mechanical axis is the gold standard for total knee replacement. The knee is however a complicated joint with bones and soft tissues working together to form a mobile yet stable joint. These soft tissues should be taken into account when determining the optimal implant alignment for a patient. Notably the ligaments have a significant influence on the stability of the artificial joint. At the moment, soft tissue balancing for total knee arthroplasty remains subjective and highly dependent on surgical expertise. Preoperative surgical planning has emerged as a recent trend in orthopaedic surgery. This technology is based on three-dimensional imaging data of the patient and allows the surgeon to plan his surgical approach before going into the operating room. In this PhD dissertation, the possibility to extend preoperative planning with dynamic simulations is investigated. Subject-specific simulations could provide the surgeon with a more comprehensive understanding of the functional outcome, including the behaviour of soft tissues. Many computer models of the knee exist, but validation of such models is essential before they can be applied in clinical practice. In this work, subject-specific models of four cadaveric knees were constructed from computed tomography images. The knee models were validated by comparing experimental cadaveric measures to model-based kinematics and ligament elongations. First, the native knees was mounted onto a mechanical rig to perform squatting, measuring kinematics and ligament elongations with optical markers and extensometers. Next, total knee replacement surgery was performed on each knee by inserting a posterior-stabilised implant and the same squat simulation was performed with the implanted knees. Coronal malrotation was also introduced using tibial inserts with a built-in slope. The presented model could accurately predict tibio-femoral kinematics and ligament elongations, showing excellent correlations and small root-mean-square errors in the same order of magnitude as other studies. Native knee results presented slightly higher deviations than the predictions for the implanted knee, as a consequence of some model simplifications. The model output with malaligned tibial inserts agreed well with the experimental data. A model parameter sensitivity study showed that internal-external rotation and ligament elongations were most affected by model parameter variations. Next, the feasibility of using the computational model to optimise implant position was explored. Surrogate modelling was applied to efficiently search for an optimum for each of the four specimens. Two different optimisation objectives were formulated. First, approximation of the native knee kinematics was used as a target for the optimisation routine. The implanted kinematics after optimisation resembled closely the native knee, however, the obtained optimal implant positions were unrealistic. Secondly, the aim was to approximate native knee ligament elongations. Optimisation towards native ligament elongations showed excellent resemblance of the ligament behaviour with only limited implant position adjustments. The integration of the presented robust optimisation method into the preoperative plan was demonstrated to be feasible. We conclude that the presented modelling approach is a promising candidate for allowing surgeons to evaluate the patient-specific implant alignment. The current method provides a solid starting point to enhance the integration of computational models into clinical practice and evolve towards an optimised approach for each patient.status: publishe