Associations Between Deep Learning-Derived Infrapatellar Fat Pad Volume on MRI with Cartilage Thickness and Imaging Features of Knee Osteoarthritis: Data from the Osteoarthritis Initiative.
Authors
Affiliations (10)
Affiliations (10)
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Epidemiology and Biostatistics, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Internal Medicine, U.C. Davis Health, Sacramento, California, USA. Electronic address: [email protected].
- Department of Epidemiology and Biostatistics, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Epidemiology and Biostatistics, University of California, San Francisco, USA. Electronic address: [email protected].
- Department of Radiology and Biomedical Imaging, University of California, San Francisco, USA.
Abstract
To investigate associations between infrapatellar fat pad volume (IPFP<sub>Vol</sub>) measured on MRI with imaging features of knee osteoarthritis. This cross-sectional study used baseline data from the Osteoarthritis Initiative. IPFP<sub>Vol</sub> was quantified using a deep learning algorithm for automatic volumetric IPFP segmentation. Outcomes by subregion included cartilage thickness in the lateral tibia (LT), medial tibia (MT), lateral femur (LF), medial femur (MF), and patella, along with Whole-Organ Magnetic Resonance Imaging Score (WORMS) scores for cartilage, bone marrow edema-like lesions (BMELL), and menisci across the same subregions and the trochlea. Linear regression models included an interaction term between IPFP<sub>Vol</sub> and subregion. When interactions were significant, subregion-specific associations were estimated. Models were adjusted for relevant confounders. A total of 4,257 participants were included in the final analysis. Significant interactions between IPFP<sub>Vol</sub> and subregion were observed for cartilage thickness and all three WORMS scores (all p<0.001). Larger IPFP<sub>Vol</sub> was associated with greater cartilage thickness in the patella (β = 0.011 mm per cm³; 95% CI, 0.006 to 0.016) and LT (β = 0.015 mm per cm³; 95% CI, 0.010 to 0.019). For WORMS, larger IPFP<sub>Vol</sub> was associated with lower cartilage lesion scores in the patella (β = -0.036 points per cm³; 95% CI, -0.061 to -0.010), LF (β = -0.058 points per cm³; 95% CI, -0.076 to -0.039), and LT (β = -0.079 points per cm³; 95% CI, -0.099 to -0.060), along with lower BMELL scores in the LT (β = -0.027 points per cm³; 95% CI, -0.036 to -0.018) and lower lateral meniscus scores (β = -0.054 points per cm³; 95% CI, -0.070 to -0.037). Larger IPFP<sub>Vol</sub> was associated with greater patellofemoral cartilage thickness and less cartilage damage on WORMS scores in the patellofemoral and lateral tibiofemoral joint. These findings highlight the subregion-specific relationships between IPFP volume and knee osteoarthritis features.