Prediction of the Serial Alignment Change after Opening-Wedge High Tibial Osteotomy Based on Coronal Plane Alignment of the Knee Using Machine Learning Algorithm
The Journal of Knee Surgery · 01.07.2025 · 38 (08) · 422-431
DOI: 10.1055/a-2525-4622 ↗Kurz gefasst
Fachartikel in der Zeitschrift The Journal of Knee Surgery, erschienen am 01.07.2025. Verfasst von Joon Hee Cho und 4 weiteren Autor:innen. Beteiligte Einrichtung: Department of Orthopedic Surgery, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si,…. Hauptthemen laut MEDLINE: Machine Learning, Osteotomy, Tibia, Osteoarthritis, Knee. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 28 Literaturangaben.
Abstract (Originalsprache)
Categorization of alignment into phenotypes can be useful for predicting and analyzing postoperative alignment changes after opening-wedge high tibial osteotomy (OWHTO). The purposes of this study were to (1) develop a machine learning model for the predicting the Coronal Plane Alignment of the Knee (CPAK) phenotypes of final alignment after OWHTO, and (2) analyze predictive factors for final alignment phenotypes. Data were retrospectively collected from 163 knees that underwent OWHTO between March 2014 and December 2019. Each data were assessed at three time points: preoperatively, at 3 months postoperatively, and the final follow-up. Constitutional alignment was also evaluated. Machine learning models were developed using two independent feature sets consisting of serial radiologic parameters and CPAK phenotypes. The area under the receiver–operating characteristic curve (AUC) was used as a primary metric to determine the best model. To evaluate the feature importance, Shapley additive explanation (SHAP) analysis was also performed on the best model. Multilayer perceptron (MLP) was the best prediction model, with the highest AUC of 0.867 based on radiologic parameters and 0.783 based on CPAK phenotypes. Joint line obliquity (JLO) at 3 months postoperatively was the most important factor among the radiologic parameters for predicting the final CPAK phenotypes. The features of constitutional and preoperative alignments also contributed, although the features of alignments at 3 months postoperatively were the highest contributing predictors. In conclusion, the developed machine learning models of the MLP showed excellent performance in predicting the final CPAK phenotypes after OWHTO. Postoperative JLO was the most important radiologic parameter for predicting the final alignment. The combination of features of the constitutional, preoperative, and postoperative periods enabled high accuracy and performance in predicting the final alignment. A retrospective cohort study with the level of evidence as level III.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Autor:innen und Institutionen
- Department of Orthopedic Surgery, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Bundang-gu, Seongnam-si, Gyeonggi-do, South Korea.
Zitieren
APA
Cho, J. H., Nam, H. S., Park, S. Y., Ho, J. P. Y., & Lee, Y. S. (2025). Prediction of the Serial Alignment Change after Opening-Wedge High Tibial Osteotomy Based on Coronal Plane Alignment of the Knee Using Machine Learning Algorithm. The Journal of Knee Surgery, 38(08), 422-431. https://doi.org/10.1055/a-2525-4622
Vancouver
Cho JH, Nam HS, Park SY, Ho JPY, Lee YS. Prediction of the Serial Alignment Change after Opening-Wedge High Tibial Osteotomy Based on Coronal Plane Alignment of the Knee Using Machine Learning Algorithm. The Journal of Knee Surgery. 2025;38(08):422-431. doi:10.1055/a-2525-4622
BibTeX
@article{Cho2025,
author = {Cho, Joon Hee and Nam, Hee Seung and Park, Seong Yun and Ho, Jade Pei Yuik and Lee, Yong Seuk},
title = {Prediction of the Serial Alignment Change after Opening-Wedge High Tibial Osteotomy Based on Coronal Plane Alignment of the Knee Using Machine Learning Algorithm},
journal = {The Journal of Knee Surgery},
year = {2025},
volume = {38},
number = {08},
pages = {422--431},
doi = {10.1055/a-2525-4622}
}
EUVIMED