University of Twente Student Theses
AI-Accelerated Prediction of Optimal Implant Alignment in Total Knee Arthroplasty
Klooster, L.R. ten (2025) AI-Accelerated Prediction of Optimal Implant Alignment in Total Knee Arthroplasty.
Full text not available from this repository.
| Full Text Status: | Access to this publication is restricted |
| Embargo date: | 1 January 2027 |
| Abstract: | Total knee arthroplasty (TKA) is effective for treating end-stage knee osteoarthritis (OA). However, dissatisfaction for one in five patients persists. A musculoskeletal model-based approach has been previously developed to predict optimal implant positioning to restore pre-diseased knee functioning and improve patient satisfaction. However, its computational complexity limits clinical application. The study aims to accelerate optimal implant alignment estimation using an AI-based surrogate model. A dataset of knee OA patients, including implant positions and their deviations in kinematics and ligament strains from the pre-diseased state, was used to train neural networks. Models tested included a multilayer perceptron (MLP), quaternion network, and multi-task MLP. These were trained per patient to predict implant position quality based on kinematics and ligament strains. For nine patients, the networks’ ability to predict implant quality was assessed. A gradient-based method was then applied to identify optimal implant positions. A conditioned network incorporating a patient identifier (hip-knee-ankle angle) was also developed to generalize across patients and tested for five cases. Strong correlations between predicted and computed values were observed for most ligaments and kinematics, though weaker for some. The quaternion network’s complexity did not improve performance over the MLP, while the multi-task MLP outperformed the single-task MLP. Gradient-based optimization identified either a single or multiple optimal positions per patient. For the conditioned network, results show that the current implementation is not sufficient for the network to distinguish between patients and generalize predictions of implant position quality. Using neural networks as surrogate models substantially accelerates the prediction of optimal implant positioning in TKA. The multi-task MLP demonstrated superior performance, while the conditioned network requires further refinement to generalize across patients. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 31 mathematics |
| Programme: | Applied Mathematics MSc (60348) |
| Link to this item: | https://purl.utwente.nl/essays/105025 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page