Computationally Efficient Optimization Method to Quantify the Required Surgical Accuracy for a Ligament Balanced TKA.

Laura Bartsoen,Matthias G.R Faes,Jos Vander Sloten,David Moens,Roel Wirix-Speetjens,Ilse Jonkers,Mariska Wesseling

doi:10.1109/tbme.2021.3069330

Laura Bartsoen, Matthias G.R Faes + Show 5 more

Open Access

https://doi.org/10.1109/tbme.2021.3069330

Copy DOI

Abstract

This study proposes a computationally efficient method to quantify the effect of surgical inaccuracies on ligament strain in total knee arthroplasty (TKA). More specifically, this study describes a framework to determine the implant position and required surgical accuracy that results in a ligament balanced post-operative outcome with a probability of 90%. The response surface method is used to translate uncertainty in the implant position parameters to uncertainty in the ligament strain. The designed uncertainty quantification technique allows for an optimization with feasible computational cost towards the planned implant position and the tolerated surgical error for each of the twelve degrees of freedom of the implant position. It is shown that the error does not allow for a ligament balanced TKA with a probability of 90% using preoperative planning. Six critical implant position parameters can be identified, namely AP translation, PD translation, VV rotation, IE rotation for the femoral component and PD translation, VV rotation for the tibial component. We introduced an optimization process that allows for the computation of the required surgical accuracy for a ligament balanced postoperative outcome using preoperative planning with feasible computational cost. Towards the research society, the proposed method allows for a computationally efficient uncertainty quantification on a complex model. Towards surgical technique developers, six critical implant position parameters were identified, which should be the focus when refining surgical accuracy of TKA, leveraging better patient satisfaction.

Highlights

A FTER a total knee arthroplasty (TKA), 20-30% of patients suffer from persisting pain, joint stiffness and/orManuscript received ??; accepted ??
For the Poly model, the decrease in error with increasing amount of samples is very small, as a Poly model is not capable of explaining higher order elements (> 2) in the musculoskeletal model (MSM). For both support vector regression (SVR) and Guassian process regression (GPR), the errors keep decreasing but at 3000 samples they are still considerably larger compared to artificial neural networks (ANN) at 1000 samples
T HE designed optimization process allows for the computation of the required surgical accuracy for a ligament balanced post-operative outcome using preoperative planning with a feasible computational cost

Summary

Introduction

A FTER a total knee arthroplasty (TKA), 20-30% of patients suffer from persisting pain, joint stiffness and/orManuscript received ??; accepted ??. Most preoperative planning processes solely account for bone geometry when determining an implant position that is consistent with a mechanically aligned TKA. Mechanical alignment does not account for strain in the ligaments, whereas several studies [4]–[6] describe that not accounting for balanced ligament strain when determining the ideal implant position is the cause of different failure types and the high patient dissatisfaction. This bound is still below the ultimate tensile strain of a human knee ligament [9] Such a ligament balance evaluation in a preoperative planning step requires complex modeling in order to predict the post-operative strain in the ligaments during movement. To determine the required surgical accuracy and account for the associated uncertainty level, the effect of surgical inaccuracies on the postoperative implant position and the associated risk of ligament imbalance and post-operative outcome

Objectives

Methods

Results

Discussion

Conclusion