Background/Purpose: Diagnosis and monitoring of patients with rheumatoid arthritis (RA) is based on radiographs or MRI, therefore the sensitivity to detect small changes of bone erosions is limited. We developed a fully 3D automated method for precise quantification of longitudinal changes of erosion size and shape in high resolution peripheral quantitative computed tomography (HR-pQCT) images of the metacarpophalangeal (MCP) joints.

Methods: HR-pQCT images of the second to fourth MCP joint from 17 RA patients were acquired on an XtremeCT scanner (Scanco, Switzerland) with an isotropic voxel size of 82μm. Patients were scanned at baseline and after 1.1±0.6 years. In order to reduce precision errors baseline (BL) and follow-up (FU) datasets were 3D rigidly registered using the periosteal surfaces to ensure identical segmentation of the cortical breaks. Erosions were then segmented independently in both BL and FU data. We evaluated change in volume, surface area, and sphericity (a measure between 0 and 1 for how closely the erosion resembles a sphere). We also dilated each erosion surface to obtain 4 concentric volumes of interest (VOIs 1-4, see figure), in which BMD was measured. The patients were treated using either Methotrexate (MTX, N=10, mean age±SD 55.8±19 years, 2 males) or tumor necrosis factor inhibitors (TNFi, N=7, 45.5±12.8 years, 1 male) and group differences were also analyzed.

Results: 37 Erosions were analyzed in both BL and FU. % change in erosion volume correlated significantly (p<0.01) in all 4 VOIs with % change in BMD. Spearman's rho were -0.58 (VOI1), -0.64 (VOI2), -0.62 (VOI3) and -0.45 (VOI4). % differences (Mean±SD) between the two treatment groups were 14.3±51.3, 13.8±44.7 and -2.7±10 in the MTX group (21 erosions) for volume, surface area and sphericity, respectively, and -11±31.5, -8.9±27.6 and 3.9±17.9 in the TNFi group (16 erosions). % BMD changes in VOIs 1 to 4 were -0.6±11.2, 0.6±12.2, -0.1±9.8 and -0.6±7.8 in the MTX group and 7.7±18.5, 10.7±23.5, 8.6±16.8 and 5.8±12.4 in the TNFi group. These differences were not significant, probably due to the small sample size.

Conclusion: We developed an automated, precise and fully 3D framework for longitudinal evaluation of changes in bone erosions in HR-pQCT images. Under treatment changes in erosion volume negatively correlated with BMD changes in the vicinity of the erosion. The loss of periarticular bone mass as a sign of inflammatory activity in RA has already been suggested in radiographic studies; interestingly, BMD did not further decrease (MTX) or increased (TNFi) after one year. However, it is still unclear, whether this is of predictive value for monitoring of disease activity. A limitation of the HR-pQCT technique are motion artifacts due to long scan times (~8min), impairing image quality.