Background

Longitudinal assessment of bone marrow lesions (BMLs) in knee osteoarthritis (KOA) by MRI is usually performed using semi-quantitative grading methods. Quantitative segmentation methods may be more sensitive to detect change over time. The purpose of this study was to evaluate and compare the validity and sensitivity to detect changes of two quantitative MR segmentation methods for measuring BMLs in KOA, one computer assisted (CAS) and one manual (MS) method.

METHODS

Twenty-two patients with KOA confined to the medial femoro-tibial compartment obtained MRI at baseline and follow-up (median 334 days in between). STIR, T1 and fat saturated T1 post-contrast sequences were obtained using a 1.5 T system. The 44 sagittal STIR sequences were assessed independently by two readers for quantification of BML. The signal intensities (SIs) of the normal bone marrow in the lateral femoral condyles and tibial plateaus were used as threshold values. The volume of bone marrow with SIs exceeding the threshold values (BML) was measured in the medial femoral condyle and tibial plateau and related to the total volume of the condyles/plateaus.The 95% limits of agreement at baseline were used to determine the sensitivity to change.

RESULTS

The mean threshold values of CAS and MS were almost identical but the absolute and relative BML volumes differed being 1319 mm3/10% and 1828 mm3/15% in the femur and 941 mm3/7% and 2097 mm3/18% in the tibia using CAS and MS, respectively. The BML volumes obtained by CAS and MS were significantly correlated but the tissue changes measured were different. The volume of voxels exceeding the threshold values was measured by CAS whereas MS included intervening voxels with normal SI.The 95% limits of agreement were narrower by CAS than by MS; a significant change of relative BML by CAS was outside the limits of -2.0%-4.7% whereas the limits by MS were -6.9%-8.2%. The BML changed significantly in 13 knees using CAS and in 10 knees by MS.

CONCLUSION

CAS was a reliable method for measuring BML and more sensitive to detect changes over time than MS. The BML volumes measured by the two methods differed but were significantly correlated.