Cartilage cavity-an MRI marker of cartilage lesions in knee OA with Data from CCBR, OAI, and PROOF

Purpose: We introduce the quantitative, continuous marker cartilage cavity that quantifies cartilage lesions by the total lesion volume. The aim was to quantify small lesions as well as large, full-depth lesions.
Methods: We included 315 knees from the Center for Clinical and Basic Research (CCBR), 972 knees from the Osteoarthritis Initiative (OAI), and 791 knees from the Prevention of OA in Overweight Females (PROOF) cohorts. In a subset, we digitally inserted artificial lesions. Each knee MRI was segmented using the knee imaging quantification (KIQ) framework. We quantified cartilage mean thickness and cavity from high-resolution cartilage thickness maps. Finally, we quantified lesion volume by the gradient peak method (GPM).
Results: Scan-rescan precision for cartilage cavity was 7.1%/3.0%. The cartilage cavity accuracy on the artificial lesions was determined as linear correlation at 0.88 with an average 8% under-estimation of lesion volume. Cavity and degree of radiographic osteoarthritis (ROA) correlated for all compartments (Spearman's rho between 0.14-0.56, P<0.001). Cavity had modest correlations to whole-organ magnetic resonance imaging score (WORMS) cartilage lesion scores but strong correlations with Boston-Leeds osteoarthritis knee score (BLOKS)/MRI osteoarthritis knee score (MOAKS) scores in most compartments (rho between 0.08-0.65, P<0.001). Cavity correlated with WOMAC pain for all tibio-femoral compartments in OAI (rho between 0.19-0.25, P<0.001) and most compartments in PROOF. Comparing with the GPM estimate, cavity was more precise, more accurate, and correlated stronger with ROA, lesion scores, and pain levels.
Conclusion: The strong correlations with ROA, radiologist lesion scores, and pain demonstrated that cavity captured OA and lesion features. Thereby, it may be appropriate for quantification of cartilage surface irregularity.