Deep contrastive learning improves identification of early-stage knee osteoarthritis across multicohort X-ray datasets.
Authors
Affiliations (15)
Affiliations (15)
- Department of Orthopaedic Surgery, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, China.
- School of Computer Science and Technology, Xinjiang University, Urumchi, China.
- Department of Health Statistics, School of Public Health, Sun Yat-sen University, Guangzhou, China.
- Department of Orthopaedic Surgery, Shanxi Medical University Second Affiliated Hospital, Taiyuan, China.
- Department of Emergency Surgery, The Affiliated Hospital of Guizhou Medical University, Guiyang, China.
- Department of Orthopedics, Orthopedic Research Institute, West China Hospital, Sichuan University, Chengdu, Sichuan Province, China.
- Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA.
- Orthopedic Sports Medicine Center, Beijing Tsinghua Changgung Hospital, Affiliated Hospital of Tsinghua University, Beijing, China.
- Department of Oncology, Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University, Nanjing, China.
- State Key Laboratory of Pharmaceutical Biotechnology, School of Life Sciences, Nanjing University, Nanjing, China.
- Shanxi Medical University, Taiyuan, China.
- Zhejiang University School of Medicine, Zhejiang University, Hangzhou, China.
- School of Biomedical Engineering, Faculty of Engineering and IT, University of Technology Sydney, Sydney, New South Wales, Australia.
- Harvard Medical School, Boston, Massachusetts, USA.
- Arthritis Clinical and Research Centre, Peking University People's Hospital, Beijing, China.
Abstract
To develop a Kellgren-Lawrence (K-L) grading recognition framework for knee osteoarthritis (KOA) with enhanced capability for early-stage detection and to validate its transferability across three independent cohorts. Weight-bearing anteroposterior knee radiographs were obtained from three datasets: the osteoarthritis initiative (OAI), Wuchuan and Shunyi. The OAI dataset included baseline, 72-month, and 96-month follow-up images, while the Wuchuan and Shunyi datasets were collected from Wuchuan (China) and Shunyi District (Beijing), respectively. Contrastive learning was incorporated into model training to construct the Augmented Dataset-Wide-ResMRnet-Contrastive Loss-Cross Entropy (AW2C) framework. The AW2C framework achieved overall classification accuracies of 83.0%, 82.0% and 80.5% on the OAI, Wuchuan and Shunyi datasets, respectively, with corresponding area under the curve (AUC) of 97.0%, 96.7% and 95.6%. Compared with the baseline model, accuracy for K-L grade 2 improved from 64% to 80%, and discrimination between K-L grades 1 and 2 was notably enhanced. The proposed AW2C framework demonstrated robust and transferable performance for automated radiographic K-L grading of KOA, particularly improving recognition of early-stage and suspected disease. With further optimisation, it holds promise as a reliable tool for large-scale studies and clinical decision support. Level III.