Predicting Total Knee Arthroplasty (TKA) Outcomes with Patient-Specific Modeling

Posted on 31 January 2022 by Kerim Genc


Computational simulation of knee movement enables Total Knee Arthroplasty (TKA) outcomes to be studied via patient-specific models based on post-operative TKA joint dynamics. 360MedCare use Simpleware software to build these models, and have recently developed a method for evaluating surgical results using a combination of 3D image processing/analysis and a predictive algorithm to relate patient-reported outcomes to the outputs of a computational simulation of a deep knee bend. Engineers from 360MedCare worked with a range of hospitals and orthopedic research institutes and companies to obtain new insights into TKA planning and recently published their findings in the journal The Knee.

The process of 3D segmentation through to patient-specific simulation generation (first cohort within this study).

Developing a Pre-Surgical Tool

Although TKA is a generally successful procedure, some patients experience post-operative problems, with poorly positioned implants being one of the main causes. Improving our understanding of knee dynamics for specific patients helps, and how they interact with implants is important for creating data to improve pre-surgical planning. Virtual preoperative trials of the resulting kinematics of a given implant selection and component alignment can look at simulation and comparison of different alignment plans, which when linked to patient-reported outcomes (PROs), might allow a predictive algorithm to select plans for future patients.

The full workflow is briefly summarized below:

  • Image data on different TKA patients was collected using a computed tomography (CT) protocol, including a cohort with pre-and-post-surgical scans and outcome surveys, and a randomly chosen sample of knees with pre-operative scans but not necessarily post-operative scans
  • 3D reconstructed patient femora and tibia were generated from the image data in Simpleware software using a semi-automated process, with anatomical landmarks placed to capture patient-specific soft tissue attachments and bone axes
  • For the first cohort, pre-operative bones and 3D implant files were registered to post-operative CT scans segmented using Simpleware software following TKA
  • Simulations were run on the models for a multibody simulation of a deep knee bend, and results used in a machine learning algorithm run to predict attainment of a Patient Acceptable Symptom State (PASS) score in captured 12-month post-operative Knee Injury and Osteoarthritis Outcome Scores (KOSS)
  • For the second cohort, pre-operative patient data was used with the predictive model to assess its performance for clinical planning
  • The generated predictive algorithm (Dynamic Knee Score) contained different features, with the Area Under the Curve (AOC) for predicting attainment of the PASS KOOS score being 0.64

Patient-specific simulation in early flexion during the extension (active) portion of the cycle.

Future Impact

The study found that the method has potential as a powerful additional tool to TKA surgical decision-making processes, but needs to be used in addition to existing techniques and within the context of non-surgical factors not covered by the project. From this approach, applications might include helping to select between otherwise reasonable surgical alignment plans to identify optimized outcomes. Considered alongside the growing use of image-based surgical planning workflows for the hip and other anatomies, as well as the development of Machine Learning-based approaches to speeding up common tasks, 360MedCare's work shows strong promise for future applications and we look forward to seeing their progress as they continue to grow.

Any Questions?

If you would like to know more, please contact us. Our technical specialists will be happy to show you Simpleware software and our Machine Learning-based solutions for any anatomy.