Fig 1.
Algorithm for image analysis of erosion parameters for metacarpophalangeal (MCP) joints.
A. Gray-scale image of the metacarpal head in the axial view. B. Binarization at a threshold value of 320 mg/cm3. C. Filling bone marrow spaces. D. Filling concavities on bone surface. E. Shrinking the contour from Fig 1D by 0.25 mm. F. Extraction of all concavities (white arrows) and detection of erosions based on the false-positive pattern in Fig 2 and the definition from SPECTRA. G. Manual measurement of erosion depth (Er.D) and width (Er.W) in the axial view. H. Manual measurement of erosion depth (Er.D-2) and length (Er.L) in the perpendicular (coronal) view. I. Measurement of peripheral BMD (Er.BMD) surrounding an erosion at a width of 1 mm. J. 3D image of MCH and erosion on the radial side (black arrow). K. Axial section of the 3D image including an erosion. L. Volume of interest (VOI) of an erosion and measurement of erosion volume (Er.V).
Fig 2.
Typical false-positive patterns of erosions in our algorithm.
A. Physiological concavity between tubercles (white arrows: tubercles). B. Pseudo-concavity formed by osteophytes (white arrows: osteophytes). C. Small cortical interruption or vascular channel (white arrow: vascular channel).
Fig 3.
Additional procedures required for some erosions in our algorithm.
A. This step may be necessary if the erosion VOI overflows beyond the concavity. a, b) These images represent overflow of the VOI (white arrow) on 2D (a) and 3D (b) images. c, d) The appropriate VOI is created by adding shrinking operations at the step in Fig 1E. B. These steps may be necessary if the erosion VOI connects with bone marrow spaces. a) The primary VOI is connected with the bone marrow spaces. b) The shrinking operation is executed iteratively until the continuity is completely interrupted. c) The dilation operation is executed the same number of times until a return to the original shape is achieved. d) Only the VOI of the erosion is extracted. C. Adjustment of rotation in the axial view may be necessary if the rotation is markedly different in longitudinal analysis. a, c) These images corresponding to the image in Fig 1D are created from the same image. A different rotation causes a slight difference in filling concavities. b, d) The two VOIs of erosions are slightly different due to differences in rotation.
Fig 4.
2D images of a metacarpal head in a healthy woman obtained from three scans with repositioning (white arrows: Erosion).
A. 1st scan; B. 2nd scan; C. 3rd scan.
Table 1.
Measurement reproducibility of each erosion parameter.
Table 2.
Characteristics of RA patients (n = 26).
Table 3.
Longitudinal changes in erosion parameters for patients with RA (n = 26).
Fig 5.
Cumulative probability plots regarding longitudinal changes in each erosion parameter for patients with RA.
Each absolute least significant change (LSCSD) is shown as a broken line. Plots exceeding the LSC indicate significant changes on both positive and negative sides. (Er.V, erosion volume; Er.D, erosion depth in the axial view; Er.D-2, erosion depth in the perpendicular (coronal or sagittal) view; Er.W, erosion width in the axial view; Er.L, erosion length in the perpendicular view; Er.BMD, peripheral bone mineral density surrounding an erosion).
Fig 6.
Two representative images of a longitudinal erosion change in patients with RA.
Each 2D MCH image represents an image at baseline (0 months) and at follow-up (12 months) from both the axial view (upper) and the coronal view (lower). A. New erosion is evident in a 74-year-old patient (white arrows indicate new erosion). Clinical data at baseline were as follows: treatment with tocilizumab; disease duration, 60 months; rheumatoid factor (RF), positive; C-reactive protein (CRP), 0.89 mg/L; Disease Activity Score in 28 joints (DAS28)-CRP, 6.70. B. Partial repair of erosion is evident in a 49-year-old patient (white arrows indicate new bone formation). Clinical data at baseline were as follows: treatment with adalimumab; disease duration, 24 months; RF, negative; CRP, 1.42 mg/L; DAS28-CRP, 3.95.