Deep learning-based effusion-synovitis volume measured on MRI is associated with osteoarthritis progression: A longitudinal analysis of data from the osteoarthritis initiative.
Authors
Affiliations (5)
Affiliations (5)
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA.
- Department of Radiology, Balgrist University Hospital, Zurich, Switzerland.
- Faculty of Medicine, University of Zurich, Zurich, Switzerland.
- Department of Epidemiology and Biostatistics, University of California, San Francisco, USA.
- Department of Internal Medicine, U.C. Davis Health, Sacramento, CA, USA.
Abstract
To investigate whether the 48-month change in effusion-synovitis volume (ΔESV) is associated with concurrent knee osteoarthritis progression, and to compare these associations with those of semiquantitative change in effusion-synovitis using the MRI Osteoarthritis Knee Score (ΔMOAKS). In this study using data from the Osteoarthritis Initiative conducted from 02/2004-10/2015, deep learning-based measurements of knee ESV were derived from baseline and 48-month follow-up knee MRI (n = 2469). OA outcomes included change of Kellgren-Lawrence (KL) grade, Whole-Organ Magnetic Resonance Imaging Score (WORMS) and its subscales (meniscus, bone marrow edema-like lesions [BMELL], cartilage), and symptom progression by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Associations of ESV change (ΔESV) with OA outcome changes were evaluated using spline-based regression models. Effect sizes were reported as interdecile differences (IDDs; 90th vs 10th percentile of the exposure). To compare the associations between ΔESV and ΔMOAKS, differences in IDD (ΔIDD) were estimated. ΔESV was significantly associated with ΔKL grade, ΔWORMS<sub>Total</sub>, ΔWORMS<sub>Meniscus</sub>, ΔWORMS<sub>BMELL</sub>, ΔWORMS<sub>Cartilage</sub>, and ΔWOMAC (all p < 0.001). Compared with ΔMOAKS, ΔESV demonstrated significantly stronger associations for ΔKL grade (ΔIDD = 0.22; p = 0.002), ΔWORMS<sub>Total</sub> (ΔIDD = 0.21; p = 0.008), and ΔWORMS<sub>Meniscus</sub> (ΔIDD = 0.20; p = 0.026), whereas differences were non-significant for ΔWORMS<sub>BMELL</sub> (p = 0.51), ΔWORMS<sub>Cartilage</sub> (p = 0.25), or ΔWOMAC (p = 0.19). Longitudinal change in ESV was associated with concurrent imaging-based and symptomatic osteoarthritis progression over 48 months. Compared with changes in MOAKS scores, ΔESV showed stronger associations with imaging-based OA outcomes. These findings underscore the potential of MRI-based ΔESV as an osteoarthritis imaging biomarker, while further studies are needed to establish its predictive and clinical utility.