2.1

This was a cross-sectional IRB-approved study (ID #201904825) where patients signed a written informed consent. We enrolled healthy volunteers with no history of prior injury or surgeries in the foot and ankle. Inclusion criteria comprised of patients 18–70 years of age at the time of the WBCT scan. Each volunteer underwent bilateral WBCT imaging of the foot and ankle. The WBCT scan of each patient was evaluated at 4 joint sites (TT, ST, TN, and first MTP). Joints with no clinical or radiographic signs of OA or other bony abnormalities were selected to be included in the study for computational analysis of JSW using a HU-based algorithm . We excluded from the analysis patients with foot deformities (cavovarus, flatfoot, hallux valgus). A total of 20 volunteers (40 feet, 9 females/11 males) were initially enrolled. Three patients were excluded secondary to recognition of the presence of foot deformities. After excluding from analysis joints that had minor osteophytes or any other early OA signs, we finally analyzed 30 TT joints (13 L, 17 R), 28 ST joints (12 R, 16 L), 26 TN joints (11 R, 15 L), and 30 first MTP joints (14 R, 16 L). The average age and body mass index (BMI) for patients analyzed can be found in Table 1 .

2.2

All volunteers were imaged using a cone-beam WBCT scanner (HiRise®, LLC, Warrington, PA, USA) with a kVp = 120 and an mA = 5. Images were acquired bilaterally while patients stood using full weight bearing capabilities and both feet pointing forward, shoulder-width apart, dividing the body weight evenly in both lower extremities. Images were reconstructed into 533 slices using 0.37 mm voxels with the reconstruction software provided with the scanner .

2.3

When analyzing WBCT images, we used the method previously described by Tazegul et al. . CubeVue® software (CurveBeam LLC, Warrington, PA, USA), was used to reconstruct the images and present the 3D images in axial, sagittal, and coronal planes that could be adjusted.

Axial, sagittal, and coronal planes were defined by first positioning the multiplanar cross-sectional tool at the center of each joint, using midpoint linear measurements in the coronal and sagittal planes of the talar dome, talar posterior facet, navicular articular surface, and articular surface of the first toe proximal phalanx for the TT, TN, ST, and first MTP joints, respectively ( Fig. 1 ). The sagittal and coronal planes were selected by positioning crosshairs at the midpoint of the anteroposterior and mediolateral portion of the articular surface of the joint analyzed. Once the crosshairs were positioned in the center of the joint, they were rotated accordingly to ensure that the axial plane was always parallel to the articular surface of each of the joints.

Sagittal (A), coronal (B), and axial (C) plane positioning of the multiplanar cross-section in the central aspect of the tibiotalar, subtalar, talonavicular and first metatarsophalangeal joints.

Once the crosshairs were placed at the center of the joint space, the volume of interest (VOI) function of the software was used. This tool generated a 3D cube with a predefined VOI whose center was positioned in the crosshair location previously placed. For the TT, ST, TN, and first MTP joints, the VOI cube dimensions were 20x20x20mm, 20x20x20mm, 15x15x15mm, and 10x10x10mm, respectively. This software tool allowed the assessment of voxel intensity distributions within the selected VOI.

Once the VOI tool was selected, a HU heatmap automatically opened in a separate window of the software to visually represent the mean intensity of the voxels in the axial plane ( Fig. 2 ). For each joint analyzed, 5 parallel linear spatial HU projections were hand-placed on the heatmap perpendicular to the axial plane, traversing the joint. One projection was placed in the center of each quadrant of the cube, and one additional projection was placed in the center of the cube. Fig. 3 shows a 3D schematic of the VOI cube placed for the TT joint with the location of the 5 projections indicated . For the TT and ST joint, the four projections were placed in the anteromedial, anterolateral, posteromedial, and posterolateral quadrants of the joint. For the TN and first MTP joints, the four projections were placed in the superomedial, superolateral, inferomedial, and inferolateral quadrants.

Heat map (A) and graphical plots (B) of Hounsfield Unit (HU) intensity profiles for each projection placed inside the cubic volume of interest (VOI). Five points marked to generate HU distribution lines. Points 1–4 marked on the center of four quadrants (anteromedial, anterolateral, posteromedial and posterolateral for the ankle joint depicted in this image) (C) and Point 5 on the central aspect of the cube (D).

Three-dimensional volume of interest (VOI) in the shape of a cube, and positioning of a total of five Hounsfield Unit lines on the central aspect of four quadrants and central aspect of the entire cube. VOI for tibiotalar and subtalar joints as depicted here, measuring 20x20x20 mm. Dimensions for the talonavicular and first metatarsophalangeal joints were respectively 15x15x15 mm and 10x10x10 mm.

The software then generated a summary of the WBCT image HU intensities along the placed projection lines going across the joint space and graphed the intensities on a plot ( Fig. 2 B) . For example, the image intensity graph for the TT joint corresponds to the transition of distal tibia cancellous and subchondral bone, to the joint space, to talar subchondral and cancellous bone . These graphical plots of the image intensity profiles and HU distributions for each projection were then used to calculate JSW and HU contrast.

2.4

Each projection allowed for the computation of JSW and HU contrast at that specific location across each assessed joint ( Fig. 4 ). First, the two maximum intensity values (I _max1 and I _max2 ) on the graphical plot, coinciding with the opposing subchondral bone surfaces, were manually selected (two peak points on the graphical plot, Figs. 4 B and 4 C). Next, the minimum value within the same plot, corresponding to the joint space, was manually selected (I _min ), corresponding to the lowest point on the HU intensity graph ( Fig. 4 C). To calculate JSW in millimeters, we calculated the voxel distance between the two maximum points selected (I _max1 and I _max2 ), multiplied by 20/54, which corresponds to the scale of millimeters for each voxel. Each joint had 5 different JSW values in millimeters corresponding to the 4 quadrants and center of the joint.

Example of a Hounsfield Unit (HU) intensity projection line crossing the tibiotalar joint (A). Graphical plot of the HU line/projection demonstrating the transition between cancellous bone, subchondral bone (peak), joint (valley), subchondral bone (peak) and cancellous bone (B). Calculation of Joint Space Width is performed in-between the two peaks in the graphical plot (subchondral bone to subchondral bone distance). HU Contrast is calculated following the depicted mathematical formula that takes into consideration the maximum and minimum HU values in the HU projection line.

To calculate the HU Michelson Contrast, which measures the difference in image intensity between the densest subchondral bone and that within the joint space, first the I _max,avg was calculated by finding the average of I _max1 and I _max2 . Next, the following formula was used: HU Contrast = (I _max,avg – I _min )/(I _max,avg + I _min ) * 100.

2.5

Continuous data were analyzed for normality by the Shapiro-Wilk test. Descriptive statistics were utilized to describe distributions, mean/median values, standard deviations (SD) and 95 % confidence intervals (CI). JSW and HU Contrast data were then compared between each joint and among the different HU projections (four quadrants and joint center) by utilizing One-Way ANOVA followed by paired T-test analysis, or Kruskal-Wallis Rank Sum test, followed by paired-Wilcoxon depending on the distribution parameters of the data. P-values less than 0.05 were considered significant. All statistical analyses were performed using JMP Pro 16® software by an independent observer.