Related Experiment Video
Updated: May 26, 2026
06:45
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
Clinical Features Outperform MRI Radiomics for Predicting Intra-Articular Hyaluronic Acid Response in Knee
Tariq Alkhatatbeh1,2,3, Ahmad Alkhatatbeh4, Yan Liao1,2,3
1Department of Joint Surgery, Center for Orthopaedic Surgery, The Third Affiliated Hospital of Southern Medical University (Academy of Orthopedics Guangdong Province), Guangzhou, China.
Annals of the New York Academy of Sciences
|May 25, 2026
Summary
Clinical variables like pain and age better predict knee osteoarthritis treatment response than MRI radiomics. Rigorous validation is crucial, as improper methods overestimate radiomics
Area of Science:
- Musculoskeletal Imaging
- Radiomics
- Osteoarthritis Research
Background:
- Intra-articular hyaluronic acid (IAHA) injections are a common treatment for knee osteoarthritis.
- Clinical response to IAHA is highly variable, necessitating predictive biomarkers.
- Quantitative magnetic resonance imaging (MRI) radiomics is explored for predicting treatment outcomes.
Purpose of the Study:
- To evaluate if MRI radiomics provides incremental predictive value for IAHA response over clinical variables.
- To assess the impact of cross-validation strategies on model performance in predicting IAHA response.
- To compare the predictive performance of clinical-only, radiomics-only, and combined models.
Main Methods:
- Utilized data from the Osteoarthritis Initiative (OAI) including 262 participants (329 knees) treated with IAHA.
- Extracted radiomics features from baseline 3D DESS MRI scans for 106 participants (128 knees).
- Employed rigorous participant-level grouped cross-validation to compare clinical, radiomics, and combined predictive models.
Main Results:
- The clinical-only model demonstrated the highest predictive performance (ROC-AUC: 0.716 in the full cohort).
- The radiomics-only model failed to predict IAHA response (ROC-AUC: 0.499).
- The combined model did not outperform the clinical-only model (ROC-AUC: 0.693), and improper validation inflated estimates.
Conclusions:
- Routine clinical variables (baseline pain, age) are superior to DESS-MRI radiomics for predicting IAHA response.
- Improper validation strategies significantly overestimate radiomics utility in musculoskeletal imaging.
- Participant-level grouping is critical for accurate validation in radiomics research to prevent data leakage.
