Bone marrow lesions in the knee diagnosed in MR images are associated with locally increased bone mineral density measured on CT images

Purpose: While the correlation between bone marrow lesions (BML) and the progressive deterioration of hyaline cartilage is evident in patients with knee OA, the association with bone mineral density (BMD) is ambiguous. We used advanced 3D QCT imaging with a dedicated MIAF analysis software to measure BMD and CT-MR registration to correlate BML with BMD. Methods: 3D CT images of the knee of 34 patients with knee OA (KL grade 2 or 3) were analyzed with MIAF-Knee starting with an automated 3D segmentation of the periosteal and endosteal surfaces of femur and tibia. Based on anatomic coordinate systems and segmented growth plates, three subchondral VOIs with increasing distance from the joint space were defined in tibia and femur (see left figure), which were first subdivided into medial and lateral and then further into anterior, central and posterior subVOIs, in which BMD was determined. BMLs were diagnosed on coronal MR images (turbo spin echo sequence) in 17 patients (WORMS grade >= 1) in either femur or tibia (BML group). 17 patients without BMLs were used as control group. A 3D registration between CT and MR data was performed to transfer all CT subVOIs to the MR datasets (see right figure). Then, BMD differences between the VOI containing the BML and its direct neighboring VOIs were calculated. For large BMLs, the BMD of all VOIs covered by the BML were averaged. Neighbor VOIs were considered in 3 dimensions (medial-lateral, proximal-distal, anterior-posterior). Corresponding to each dataset of the BML group, the BMD difference was also calculated using the same VOIs in a randomly assigned dataset of the control group. This procedure was repeated five times. Therefore, for each patient of the BML group, the corresponding BMD differences in the control group were averages derived from 5 different control datasets. Results: Despite different resolution in CT (voxel size 0.25 x 0.25 x 0.3 mm3) and MR (voxel size 0.55 x 0.55 x 3.0 mm3) acquisitions, the registrations showed very good overlap. Due to the BMD gradient in the epiphysis, percentage BMD differences between VOIs containing lesions and their neighboring VOIs were significantly different from zero in both groups. However, differences between VOIs were significantly higher (t-test, p<0.01) in the BML group (43.0%±40.0%) compared to the control group (11.7%±22.9%). Conclusions: Despite the relative small sample size, we found in patients with knee OA that the presence of BMLs was strongly associated with increased BMD at the BML location. Increased BMD associated with BML may be an indicator of locally increased loading within the knee joint or a consequence of trabecular microfractures (impaction and callus formation), although the cause-effect relationship was not investigated. However, in our study, the BML extension was not exactly covered by the a priori (in the CT image) defined analysis VOIs. Nevertheless, these results are a promising step in utilizing multimodality imaging and advanced 3D analysis for the diagnosis of knee OA by combining the benefits of CT and MR.