Fig 1.
Flowchart of the automated image analysis method for subchondral bone partition into plate and trabecular bone and volumetric quantifications.
In the first stage (A), 3D image registration was applied to micro-CT scans of DMM tibiae for alignment to a template position, while the contralateral was co-aligned with the correspondent ipsilateral (step 1 in A). Prior compartmentalisation and volumetric analysis (B), meshes were voxelized into stacks of images (step 2 in A). Volume partition was performed based on macro-porosity differences, generating two colour-coded volumes-of-interest for each aspect of the tibial plateau (medial and lateral), green for subchondral bone plate and red for trabecular bone (step 1 in B). Using these mappings, the original volumes-of-interest were extracted for quantitative analysis of each compartment (step 2 in B). The top view of the 3D model of the tibial epiphysis shows the dimensions and location of the mappings across the load-bearing areas and the dashed line demarcates the medial and lateral aspects of the plateau.
Table 1.
Mean and standard deviation (SD), Pearson correlation coefficients, root mean square (RMS) errors and coefficients of variation (%CV) of the measurements in manually and automatically segmented subchondral bone compartments.
Fig 2.
Bland-Altman plots to determine the agreement between the measurements in automatically and manually segmented subchondral bone compartments.
The difference between measurements using automated and manual segmentation was plotted against the average for the subchondral bone plate (A) volume and (B) thickness, and trabecular (C) total volume, (D) bone volume and (E) BV/TV, (n = 24 in which the measurements in the medial and lateral aspects of the plateau for both right and left legs were pooled together). The mean (solid line) and the 95% limits of agreement (-1.96SD, +1.96SD, indicated by the area shaded in blue between the dashed lines) for each microstructural parameter are reported in the plots.
Fig 3.
Heat maps and profiles of subchondral bone thickness obtained by micro-CT over OA temporal progression.
(A) Representative pairs of contralateral (left) and DMM-operated (right) tibial top surfaces at baseline, 1-, 2-, 4- and 12-weeks post-DMM, showing progressive subchondral bone thickening in the medial compartment of DMM tibiae, where colour-coding changed from orange (~150 μm) at 1-week towards white (~300 μm) at 12-weeks. Blue arrows in DMM maps indicate the presence of medial osteophytes and dashed white lines delimit the location where medial to lateral profiles were extracted for both contralateral (B) and DMM-operated (C) tibiae. In destabilised joints (C), the peak thickness in the medial compartment was continuously increased from baseline to 20-weeks post-DMM (increment of ~135 μm, indicated by the dashed arrow) while in contralateral (B) this difference was less marked (~50 μm, indicated by the dashed arrow). The superimposition of all profiles in the lateral compartment suggested that no changes occurred in this area.
Fig 4.
Automated compartmentalisation of epiphyseal subchondral bone over OA temporal progression.
(A) Middle coronal view of a 3D model showing epiphyseal compartmentalisation into subchondral bone plate (colour-coded in green) and trabecular bone (colour-coded in red) in the medial and lateral aspects of the tibial plateau, demarcated by the dashed line. (B) Representative coronal views of the mappings (500 μm in width by 350 μm in depth) at 1-, 2-, 4-, 12- and 20-weeks post-DMM, for both contralateral and DMM tibiae.
Fig 5.
Microstructural assessment of subchondral bone compartments.
Subchondral bone plate (A) volume and (B) thickness, and trabecular (C) total volume and (D) BV/TV, (E) subchondral bone plate BMD, and (F) trabecular BMD within the automated volumes-of-interest of medial and lateral aspects of the plateau for DMM-operated and contralateral tibiae (n = 6, * denotes comparisons between measurements in the non-operated baseline (time 0) with subsequent time points post-surgery, while # denotes comparisons between DMM-operated vs. contralateral aspects of tibial plateau at the same time point, obtained by two-way ANOVA followed by post hoc multiple comparison tests using Bonferroni correction). *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001 (both * and # symbols denote the same levels of significance).
Table 2.
Mean and standard deviation (SD), Pearson correlation coefficients, root mean square (RMS) errors and coefficients of variation (%CV) of the measurements in subchondral bone compartments imaged at 5 and 10 μm/pixel resolution.
Fig 6.
Osteophyte quantification using shape comparisons based on 3D image registration.
(A) Middle coronal view of a 3D model showing registration of DMM (colour-coded in red) to contralateral (colour-coded in green) tibia leading to superimposition between structures (colour-coded in yellow). The dashed line demarcates the medial and lateral aspects of the plateau. (B) Representative coronal views of registered DMM/contralateral tibiae showing osteophytes as a medial outgrowing protrusion (highlighted by dashed boxes), which were well validated by histopathology. Progressive articular cartilage damage is indicated by arrows. (C) Osteophyte volume measure either by automated or manual segmentation (n = 5, *P<0.05, **P<0.01 and ***P<0.001, computed by one-way ANOVA followed by multiple comparison tests using Bonferroni correction to determine differences between measurements in healthy baseline and subsequent time points post-surgery). (D) Bland-Altman plot to determine the agreement between the measurements in automatically and manually segmented osteophytes. The difference between measurements using automated and manual segmentation was plotted against the average (n = 35, with the measurements of all time points upon DMM pooled together). The mean (solid line) and the 95% limits of agreement (-1.96SD, +1.96SD, indicated by the area shaded in blue between the dashed lines) are reported in the plot. (E) Articular cartilage summed histopathology scores in medial and lateral aspects of the plateau for DMM and contralateral tibiae (n = 5, **P<0.01 and ***P<0.001 by non-parametric Kruskal-Wallis tests followed by post hoc Dunn’s multiple comparisons tests to determine statistical differences between scores at 1-week post-DMM and the subsequent time points for each aspect of the plateau).
Fig 7.
Osteophyte quantification using epiphyseal volume.
(A) Top views of tibial epiphyses showing medial expansion caused by osteophytes in DMM-operated (arrows and shaded regions-of-interest) compared with an unaltered contralateral (12-weeks post-surgery). At 2-weeks, the osteophyte appeared incomplete, while at 12-weeks broaden the articular surface. (B) Epiphyseal volume over time for DMM and contralateral tibiae (n = 5, * denotes comparisons between measurements in the non-operated baseline with subsequent time points post-surgery while # denotes comparisons between DMM-operated vs. contralateral tibiae at the same time point, obtained by two-way ANOVA followed by multiple comparison tests using Bonferroni correction). #P<0.05, **P<0.01 and ****P<0.0001. (C) Difference between epiphyseal volume of DMM-operated and contralateral tibiae over time expressed as a percentage difference (n = 5, *P<0.05, **P<0.01 and ***P<0.001, computed by one-way ANOVA followed by multiple comparison tests using Bonferroni correction to determine statistical differences between measurements in healthy baseline and subsequent time points).
Table 3.
Non-parametric correlations between lesion progression in subchondral bone compartments, osteophyte formation and articular cartilage score.