Automatic morphometric cartilage quantification in the medial tibial plateau from MRI for osteoarthritis grading

SummaryObjectiveTo evaluate whether a novel, fully automatic, morphometric cartilage quantification framework is suitable for assessing level of knee osteoarthritis (OA) in clinical trials.MethodThe population was designed with a normal population and groups with varying degree of OA of both sexes and at ages from 21 to 78. Posterior–anterior X-rays were acquired in semi-flexed, load-bearing position. The radiographic signs of OA were evaluated based on the Kellgren and Lawrence score (KL) and the joint space width (JSW) was measured. Turbo 3D T1 magnetic resonance imaging (MRI) scans were acquired with resolution 0.7×0.7×0.8mm3 from a 0.18T scanner. The morphometric cartilage quantification from MRI resulted in volume, surface area, thickness and surface curvature for the medial tibial cartilage compartment. These quantifications were evaluated against JSW with respect to precision and ability to separate healthy subjects from OA subjects.ResultsThe automatic, morphometric cartilage quantifications allowed fairly precise measurements with scan–rescan coefficient of variations (CVs) in the range from 3.4% to 6.3%. All quantifications, including JSW, allowed separation of the groups of healthy and OA subjects. However, for separation of the healthy from the borderline cases (KL 0 vs KL 1), only the Cartilage Curvature quantification allowed statistically significant separation (P