3D Geometric Shape Reconstruction for Revision TKA and UKA Knees Using Gaussian Process Regression

Area of Science:

• Orthopedic Surgery
• Biomedical Engineering
• Medical Imaging

Background:

• Revision knee surgery presents challenges due to altered bone geometry and potential implant malpositioning.
• Accurate preoperative planning is crucial for successful revision total knee arthroplasty (TKA) and unicompartmental knee arthroplasty (UKA).

Purpose of the Study:

• To develop and validate a statistical method for reconstructing native femoral and tibial surface shapes for revision knee surgery.
• To assess the accuracy of this reconstruction method in simulated TKA and UKA procedures.

Main Methods:

• Utilized Gaussian Process regression to reconstruct native bone geometry from simulated TKA/UKA resections.
• Analyzed prediction errors for femoral and tibial bone cuts in 20 femurs and 20 tibias.
• Investigated the influence of bone size and patient gender on prediction accuracy.

Main Results:

• Average absolute errors for predicted resected femur bones were 1.0 ± 0.3 mm (TKA), 1.0 ± 0.3 mm (medial UKA), and 0.8 ± 0.2 mm (lateral UKA).
• Average absolute errors for predicted resected tibia bones were 1.0 ± 0.4 mm (TKA), 0.8 ± 0.2 mm (medial UKA), and 0.7 ± 0.2 mm (lateral UKA).
• Prediction accuracy for UKA was significantly better than TKA (p < 0.05), with variations observed based on bone size and gender.

Conclusions:

• The Gaussian Process regression-based reconstruction is a viable approach for planning revision knee surgeries.
• This method is particularly useful when contralateral anatomy is unavailable or compromised.
• Accurate bone shape reconstruction can improve implant selection and surgical outcomes in revision TKA/UKA.