Forecasting knee replacement surgery with deep learning: an integrated approach using routine clinical data and radiographs
File(s)
Published version
Author(s)
Type
Journal Article
Abstract
Aims: We aimed to develop and validate a practical deep learning model integrating routinely collected clinical data and knee radiographs to predict the need for knee replacement (total or partial) in patients with, or at-risk of, knee osteoarthritis, as well as the time to surgery.
Methods: Data from the Multicenter Osteoarthritis Study (MOST) and the Osteoarthritis Initiative (OAI) were utilised. The MOST dataset, comprising 3,026 patients, was the primary training and testing cohort, while the OAI dataset provided external validation. The final architecture was based on DenseNet-201, with a head that combined outputs from the radiographic analysis with routinely collected clinical data. Model evaluation used area under the receiver operating characteristic curve (AUROC).
Results: The integration of clinical and radiographic data significantly improved predictive accuracy. The combined model achieved an AUC of 0.85, outperforming models using either data source alone. External validation with the OAI dataset yielded an AUC of 0.79, confirming the model's generalisability. The AUC for surgical interventions within 40 months was 0.83 on the validation dataset, demonstrating higher predictive accuracy for earlier surgical needs.
Conclusions: This study highlights the potential of deep learning models, which integrate routine clinical and radiographic data, to predict the need for knee replacement. The robust performance and generalisability of the developed model could streamline clinical pathways and predict local demand for surgery during the next three years. This will facilitate resource planning for providers and accurate and timely access to surgical interventions for patients.
Methods: Data from the Multicenter Osteoarthritis Study (MOST) and the Osteoarthritis Initiative (OAI) were utilised. The MOST dataset, comprising 3,026 patients, was the primary training and testing cohort, while the OAI dataset provided external validation. The final architecture was based on DenseNet-201, with a head that combined outputs from the radiographic analysis with routinely collected clinical data. Model evaluation used area under the receiver operating characteristic curve (AUROC).
Results: The integration of clinical and radiographic data significantly improved predictive accuracy. The combined model achieved an AUC of 0.85, outperforming models using either data source alone. External validation with the OAI dataset yielded an AUC of 0.79, confirming the model's generalisability. The AUC for surgical interventions within 40 months was 0.83 on the validation dataset, demonstrating higher predictive accuracy for earlier surgical needs.
Conclusions: This study highlights the potential of deep learning models, which integrate routine clinical and radiographic data, to predict the need for knee replacement. The robust performance and generalisability of the developed model could streamline clinical pathways and predict local demand for surgery during the next three years. This will facilitate resource planning for providers and accurate and timely access to surgical interventions for patients.
Date Issued
2026-07-14
Date Acceptance
2026-03-30
Citation
Bone & Joint Open, 2026, 7 (7), pp.926-933
ISSN
2633-1462
Publisher
The British Editorial Society of Bone & Joint Surgery
Start Page
926
End Page
933
Journal / Book Title
Bone & Joint Open
Volume
7
Issue
7
Copyright Statement
© 2026 Musbahi et al. This article is distributed under the terms of the Creative Commons Attributions (CC BY 4.0) licence (https:// creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium or format, provided the original author and source are credited.
License URL
Identifier
10.1302/2633-1462
Publication Status
Published