Capturing deep partial-thickness cartilage loss with semiquantitative scoring: A supplemented MRI Osteoarthritis Knee Score (sMOAKS).

Objective: We propose a supplement to MOAKS (MRI Osteoarthritis Knee Score) for capturing >50% partial thickness cartilage loss on knee MRI and measure reader agreement. Design: MOAKS scores 2 severity levels of cartilage damage (any loss, full-thickness loss) within knee subregions with lesional area graded 0-3. We propose supplemented MOAKS (sMOAKS) by adding a similarly graded third level assessment for deep cartilage loss (DCL), >50% thickness, in addition to traditional MOAKS for improved granularity of partial thickness cartilage loss. Using sMOAKS, two radiologists independently graded cartilage subscores for 40 knees and rescored 20 knees. Consolidated inter-reader and intra-reader agreement was calculated with kappa values for the DCL level supplement. To measure agreement for one example of a more granular combined sMOAKS outcome, coverage probability reader agreement was calculated for a scaled cartilage damage score (CDS), a summed normalized score (0-100) with equal weighting to articular surfaces combining subregion scores. Results: DCL represented 27.8% (63/227) of partial but not full-thickness cartilage loss subregion interpretations. Pooled subregion DCL involving >10% surface area demonstrated 97.7% (ĸ=0.71) inter-reader agreement and 98.8% (ĸ=0.78) intra-reader agreement. For greatest subregion DCL size, weighted ĸ agreement was 0.73/0.82 (inter-reader/intra-reader) for articular surfaces and 0.75/0.83 for joint compartments. At 90% CDS intra-reader agreement coverage, inter-reader CDS agreement values were 83%, 82%, and 78% for surfaces, compartments, and whole joints, respectively. Conclusions: There is substantial agreement for deep cartilage loss detection using sMOAKS across varied analysis methods. Further assessment will determine when the added granularity of sMOAKS is beneficial.