Infrapatellar fat pad MRI radiomics and delta-radiomics for predicting pain progression in knee osteoarthritis.

BACKGROUND: Pain is the hallmark of knee osteoarthritis (KOA) and a major cause of disability. Identification of individuals at high risk of pain progression may facilitate timely intervention. The infrapatellar fat pad (IPFP) has been implicated in KOA-related pain. We aimed to evaluate whether baseline and longitudinal delta MRI-derived IPFP radiomic features can predict pain progression. METHODS: This study used the Foundation for the National Institutes of Health (FNIH) Osteoarthritis Biomarker Consortium dataset from the Osteoarthritis Initiative (OAI) for model training and hold-out validation, while the Pivotal OAI MRI Analyses (POMA) dataset was used for independent validation. A total of 919 knees with baseline and 24-month MRI examinations (1838 scans) were included. We extracted the radiomic features of IPFP and developed a baseline radiomics model as well as a longitudinal delta-radiomics model over two years. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC) and the area under the precision-recall curve (AUC-PRC), and compared with the clinical model and a whole-joint MRI Osteoarthritis Knee Score (MOAKS) model. RESULTS: The baseline IPFP radiomics model exhibited stable predictive performance, with AUC-ROC values of 0.731 and 0.732 in the hold-out and independent validation sets, outperforming clinical and MOAKS models. In the independent validation set, the model yielded an AUC-PRC of 0.257, substantially exceeding the positive event prevalence (0.115). Notably, the delta-radiomics model further improved performance, achieving AUC-ROC values of 0.861 and 0.806 in corresponding validation sets and an AUC-PRC of 0.434 in the independent validation set. More importantly, the IPFP radiomic score was confirmed as an independent predictor of KOA pain progression in multivariable regression. CONCLUSION: Quantitative MRI radiomics of the IPFP provide a promising tool for predicting KOA pain progression.