Back to all papers

An MRI-based radiomics framework for early identification and progression stratification in knee osteoarthritis: data from the osteoarthritis initiative.

October 31, 2025pubmed logopapers

Authors

Fu J,Mu L,Dong D,Li M,Miao Z,Huai X,Zheng Y,Zhang H

Affiliations (4)

  • Department of Radiology, The First Hospital of Jilin University, No.1 of Xinmin Street, Changchun, Jilin Province, 130021, China.
  • Linking Innovations Technology Ltd, Beijing, 100176, China.
  • Department of Sports Medicine, The First Hospital of Jilin University, Changchun, Jilin Province, 130021, China.
  • Department of Radiology, The First Hospital of Jilin University, No.1 of Xinmin Street, Changchun, Jilin Province, 130021, China. [email protected].

Abstract

To develop a cascaded machine learning model based on MRI radiomics features from cartilage and subchondral bone to predict the incidence and progression of knee osteoarthritis (KOA), thereby addressing the need for early intervention in pre-radiographic stages. The study analyzed 456 participants without radiographic OA (Kellgren-Lawrence [KL] 0-1) at baseline were selected from the Osteoarthritis Initiative (OAI) and randomly divided into training and testing cohorts (7:3). Participants underwent 3D DESS MRI of the right knee and were stratified into incident KOA and non-KOA groups based on 4-year radiographic outcomes, using 1:2 propensity score matching (PSM) to adjust for baseline confounders. Early and late progressors were further classified based on the timing of radiographic progression. Radiomic features of cartilage and subchondral bone were extracted from nnU-Net-based segmentation. Optimal features were selected through Least Absolute Shrinkage and Selection Operator (LASSO) regression and principal component analysis (PCA). A two-stage logistic regression (LR) classification framework was developed to predict KOA incidence and progression, with a cascaded LR model implemented for multi-class classification. Model performance was primarily evaluated using the area under curve (AUC), with SHapley Additive exPlanations (SHAP) for interpretability. SHAP analysis identified square_glrlm_ShortRunLowGrayLevelEmphasis and wavelet-HHL_firstorder_Minimum from subchondral bone as primary predictors. The combined cartilage and subchondral bone radiomics models demonstrated superior performance in predicting incidence (AUC: 0.985, 95% CI: 0.969-1.000) and progression (AUC: 0.738, 95% CI: 0.565-0.911) of KOA. The cascaded model achieved AUCs > 0.800 across categories, with an overall accuracy of 0.791. The MRI radiomics framework integrating cartilage and subchondral bone features, effectively predicts KOA incidence and progression, enhances individualized risk stratification, and facilitates timely clinical decisions. Clinical trial number not applicable.

Topics

Osteoarthritis, KneeMagnetic Resonance ImagingMachine LearningCartilage, ArticularKnee JointJournal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.