INFRAPATELLAR FAT PAD MRI RADIOMICS AND DELTA-RADIOMICS FOR PREDICTING PAIN PROGRESSION IN KNEE OSTEOARTHTRITIS

Knee osteoarthritis (KOA) pain is often discordant with structural abnormalities on conventional imaging, limiting the ability of routine magnetic resonance imaging (MRI) to predict symptom progression. The infrapatellar fat pad (IPFP), a metabolically active intra-articular tissue implicated in local inflammation, may provide quantitative imaging biomarkers relevant to pain progression.

To determine whether baseline IPFP MRI radiomics and two-year IPFP delta-radiomics can predict pain progression in KOA and whether their performance exceeds that of clinical features and whole-joint MRI Osteoarthritis Knee Score (MOAKS) models.

We conducted a retrospective cohort study using 1,822 knee MRI scans from the Osteoarthritis Initiative. Radiomic features were extracted from the IPFP on knee MRI. A baseline IPFP radiomics model and a longitudinal two-year delta-radiomics model were developed to predict KOA pain progression. Model performance was assessed using the area under the receiver operating characteristic curve (AUC) in internal validation and external test sets. Predictive performance was compared with models based on clinical features and MOAKS. Multivariable regression was used to evaluate whether the IPFP radiomics score was independently associated with pain progression.

The baseline IPFP radiomics model showed stable predictive performance, with AUCs of 0.731 in internal validation and 0.732 in the external test set. This exceeded the performance of clinical-feature models (AUC 0.534-0.587) and MOAKS-based models (AUC 0.532-0.586). The two-year delta-radiomics model further improved prediction, achieving an AUC of 0.861. In multivariable regression, the IPFP radiomics score remained an independent predictor of KOA pain progression.

Quantitative MRI radiomics of the IPFP may provide a robust imaging biomarker for predicting KOA pain progression. Incorporating IPFP radiomics, particularly longitudinal delta-radiomics, may improve risk stratification beyond clinical features and conventional semi-quantitative MRI assessment.