EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Cross‐Institutional Five‐Class Kellgren–Lawrence Grading of Knee Osteoarthritis via Multitask Deep Learning

Tariq Alkhatatbeh, Ahmad Alkhatatbeh, Yan Liao, Zhilin Zhang, Hang Fang, Weidong Chen, Rongkai Zhang

Annals of the New York Academy of Sciences · 2026-03 · 1557 (1)

DOI: 10.1111/nyas.70254 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Annals of the New York Academy of Sciences, online erschienen am 14.03.2026. Verfasst von Tariq Alkhatatbeh und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Joint Surgery, Center for Orthopaedic Surgery, The Third Affiliated Hospital of Southern Medical University (Academy of Orthopedics Gua…. Hauptthemen laut MEDLINE: Deep Learning, Osteoarthritis, Knee. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Deep learning models for Kellgren–Lawrence (KL) grading often report optimistic performance due to data leakage and fail to generalize across institutions because of domain shift. To address this reproducibility crisis, we introduce KL‐FuseNet, a multitask architecture fusing global (ConvNeXt‐Base) and local (ResNet‐50) features to predict ordinal grades, label distributions, and binary severity (KL≥2). Using strict patient‐wise stratified splits on an internal osteoarthritis initiative dataset ( n = 8260) and an independent Chinese cohort ( n = 2295), we compared zero‐shot transfer against selective fine‐tuning. KL‐FuseNet achieved robust internal agreement (quadratic Cohen's kappa [QWK]: 0.881; accuracy: 70.3%). While external zero‐shot deployment revealed a domain gap, with accuracy dropping to 66.1%, our selective fine‐tuning protocol significantly bridged this divide, boosting external accuracy to 80.0% and QWK to 0.950, with an AUC of 0.984 for clinically significant osteoarthritis (KL≥2). These results demonstrate that while KL‐FuseNet achieves state‐of‐the‐art performance under rigorous evaluation, domain‐aware adaptation is essential for clinical utility. This study establishes a reproducible pathway for deploying automated grading models across heterogeneous medical centers.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Deep LearningOsteoarthritis, KneeHumansReproducibility of ResultsSeverity of Illness Index

Stichwörter der Autor:innen

Kellgren–Lawrence (KL)deep learningdomain shiftknee osteoarthritis (KOA)multitask architecture

Autor:innen und Institutionen

  • Department of Joint Surgery, Center for Orthopaedic Surgery, The Third Affiliated Hospital of Southern Medical University (Academy of Orthopedics Guangdong Province), Guangzhou, China.
  • Orthopedic Hospital of Guangdong Province, Guangzhou, China.
  • Guangdong Provincial Key Laboratory of Bone and Joint Degeneration Diseases, Guangzhou, China.
  • Department of Orthopedics, The First Affiliated Hospital of Shantou University Medical College, Shantou City, China.

Zitieren

APA

Alkhatatbeh, T., Alkhatatbeh, A., Liao, Y., Zhang, Z., Fang, H., Chen, W., & Zhang, R. (2026). Cross‐Institutional Five‐Class Kellgren–Lawrence Grading of Knee Osteoarthritis via Multitask Deep Learning. Annals of the New York Academy of Sciences, 1557(1). https://doi.org/10.1111/nyas.70254

Vancouver

Alkhatatbeh T, Alkhatatbeh A, Liao Y, Zhang Z, Fang H, Chen W, et al. Cross‐Institutional Five‐Class Kellgren–Lawrence Grading of Knee Osteoarthritis via Multitask Deep Learning. Annals of the New York Academy of Sciences. 2026;1557(1). doi:10.1111/nyas.70254

BibTeX
@article{Alkhatatbeh2026,
  author = {Alkhatatbeh, Tariq and Alkhatatbeh, Ahmad and Liao, Yan and Zhang, Zhilin and Fang, Hang and Chen, Weidong and Zhang, Rongkai},
  title = {Cross‐Institutional Five‐Class Kellgren–Lawrence Grading of Knee Osteoarthritis via Multitask Deep Learning},
  journal = {Annals of the New York Academy of Sciences},
  year = {2026},
  volume = {1557},
  number = {1},
  doi = {10.1111/nyas.70254}
}
Kostenlos starten →