Comparison of the kinematics, repeatability, and reproducibility of five different multi-segment foot models

This is a cross-sectional study to analyse the differences in repeatability and reproducibility of five MFMs applied to normal healthy males.

Eleven healthy male volunteers with a mean age of 26.5 years participated in this study. The inclusion criteria were as follows: 1) no soft tissue injuries of both feet and ankles within one year from the experimental date, 2) no history of fracture or surgery on both lower extremities, 3) no abnormal findings on both feet radiographs including arthritic changes, 4) no pain during gait. The institutional review board of Seoul National University Hospital approved this study, and all participants provided informed consents prior to participation.

We selected five MFMs which have been validated and used in clinical articles [1]: DuPont foot model (DFM) [3], modified Rizzoli foot model (mRFM) [4], Oxford foot model (OFM) [5], Milwaukee foot model (MiFM) [13], and modified Shriners Hospital for Children Greenville foot model (mSHCG) [14]. We created a merged 28-marker set using nine 8-mm spheres on the shank and twenty 4-mm spheres on the foot (Fig. 1 and Additional file 1). To apply the merged set, we slightly modified the marker placements of mRFM. It attaches the markers on the second metatarsal head and base, but the other models attach these markers between the second and third metatarsal heads and bases. Since it was difficult to place them separately due to their positional difference of only a few millimeters, we placed the markers between the second and third metatarsal heads and bases only. Although hallux is an important segment to be considered during gait, it was excluded because it was too small to allow for the successful attachment of the markers of all the models.

Two physical therapists with 8 years of experience participated in attaching the merged set to all subjects. A preliminary analysis was done prior to this study to have an appropriate level of experience with other foot models in two subjects. They individually applied the merged 28-marker set to the participants and repeated it after one or two weeks to analyse inter-trial and inter-session repeatability and inter-evaluator reproducibility. To prevent offset angles from influencing foot positioning during static calibration, we maintained the participants’ feet in a fixed position during the first and second visit. Specifically, we marked two parallel straight lines on the ground using masking tape at 30 cm apart, and the center of the subject’s heel and the second toe were placed on the line. After the static calibration, the subjects walked at a self-selected speed along an 8-m walkway. Kinematic data were captured using 12 cameras with a 3D optical motion capture system (Motion Analysis Co., Santa Rosa. CA) and Cortex 1.3.0 software (Motion Analysis Co., Santa Rosa. CA) at a sampling rate of 120 Hz. Visual3D Professional v6.01.36 (C-Motion, Inc.) was used for post-processing and building the five models. According to the original paper of each MFM, the Euler/Cardan angle sequence of OFM, MiFM, and mSHCG was sagittal-coronal-transverse, while DFM and mRFM used the sagittal-transverse-coronal sequence. All kinematic data were time-normalized to 100% of the gait cycle (1% interval between time points), and three representative walking trials were chosen from each session. To use unified terminology, we defined “tibia and shank” as “shank,” “calcaneus and hindfoot” as “hindfoot,” and “metatarsus and forefoot” as “forefoot.” Then, we analysed two relative angles between three segments: shank-hindfoot (HF) and hindfoot-forefoot (FF). We indicated the tracking markers constituting the segments of each model in Additional file 1. To assist understanding of the tri-planar motions, superscripts were used: i.e., HF motion in the sagittal, coronal, and transverse planes were abbreviated as HF^sag, HF^cor, and HF^trans, respectively.

DFM, mRFM, and OFM used the general marker-based method without subtracting the static offset from the corresponding value of the walking trials, while MiFM used a method of rotating each segment in the medial/lateral, anterior/posterior, and longitudinal axes from radiographic measurements. Thus, the participants underwent three weight-bearing X-ray scans of the dorsoplantar foot, lateral view of the foot and ankle, and heel alignment view (Fig. 2). Two evaluators (physical therapists) individually collected the radiographic measurements according to the original papers’ methods and applied it to each segment rotation [18, 19]. Furthermore, mSHCG rotated the HF segment in the anterior/posterior axis according to goniometric measurements of the HF varus/valgus, and provided options of rotating the FF and HF segments to the pitch angles measured from the radiographic images, or constructing the segments with the markers only. Accordingly, we selected Option 2 for the HF, a radiograph for calcaneal pitch, and Option 3 for the FF, a marker-based FF pitch [14]. The goniometric measurement of HF valgus/varus was taken between the calcaneal axis and the line perpendicular to the floor during weight bearing standing. To reduce the effect of foot posture, we placed the participants’ feet in a fixed position similar to static calibration. Moreover, to minimize the influence of the evaluator’s angle of view during measurement, they were asked to measure the angle with their eyes at the subject’s ankle height.

Coefficients of multiple correlations (CMC) [20] and experimental errors (σ) [21] were calculated to assess inter-trial and inter-session repeatability and inter-evaluator reproducibility. Inter-trial CMC was calculated from two strides processed by one evaluator in the second session and analysed by the averaged data from three strides for each session. To determine inter-evaluator CMC, the averaged data from the second session were selected. Moreover, the inter-trial errors (σ^trial), inter-session errors (σ^sess) of each evaluator, and inter-evaluator errors (σ^eval) were calculated for each model. These were expressed as the mean standard deviation of 0 to 100 points of the whole gait cycle. The Kruskal-Wallis test and Bonferroni correction were performed to find differences in the range of motion (ROM) (α = 0.05). When we analysed CMC and ROM, SPSS statistics 25.0 for windows (SPSS Inc., Chicago, IL) was used. To compare the kinematics of five models, statistical parametric mapping (SPM) [22] of the t-value from the post-hoc unpaired t-test (α = 0.01) was employed using MATLAB (R2019a, The MathWorks, Inc., Natick, MA). SPM of the t-value was used for identifying the difference between continuous curves, which can be calculated using open-source SPM1d code (www.​spm1D.​org).

The inter-trial CMCs of HF and FF were greater than 0.886 in all models (see Table 1). In the sagittal plane, the inter-session and inter-evaluator CMCs of all segments were greater than 0.840 in all models. In HF^cor repeatability, the CMCs of DFM, OFM, and mRFM were less than 0.660, whereas those of MiFM and mSHCG were greater than 0.793. For OFM, the averaged inter-session and inter-evaluator CMCs were lower in HF^cor (0.424 and 0.410, respectively) than in HF^trans (0.810 and 0.791, respectively). Similarly, this was also observed in FF^cor (0.467 and 0.612, respectively), when compared to FF^trans (0.931 and 0.846, respectively). The repeatability and reproducibility of FF were generally greater than HF in all planes, but in FF^trans, the reproducibility of mRFM and mSHCG (0.332 and 0.441, respectively) were lower than their repeatability (0.735 and 0.735, respectively).

With respect to the averaged σ, σ^trial of all models were less than 1.3° for both HF and FF (Table 1). HF^cor, DFM, OFM, and mRFM showed σ^sess ranging from 1.8° to 2.8° and σ^eval ranging from 3.1° to 3.4°, whereas MiFM and mSHCG showed σ^sess ranging from 1.3° to 1.7° and σ^eval of 1.8° and 1.7°, respectively. FF σ^sess ranged from 0.9° to 2.6° among all models, which was generally lower than that of HF, and σ^eval ranged from 1.2° to 3.4°, which was higher than σ^sess. All averaged σ of all models did not exceed 3.5°.

Figures 3 and 4 show the mean kinematics for each model and SPM curves acquired from the post-hoc unpaired t-test of MFMs. In HF^sag, MiFM and mSHCG were more dorsiflexed than DFM, OFM, and mRFM in the whole cycle due to static offset angles, and there were no differences in ROM among MFMs (Figs. 3 and 5). In HF^cor, all SPM curves showed no significant differences except for mRFM-mSHCG, but the DFM and mSHCG had significantly greater ROMs. In HF^trans, MiFM and mSHCG showed significantly reduced ROM than the others. The SPM curves of HF for DFM-OFM were not statistically different in all planes. In contrast to HF^sag, FF^sag of MiFM, mSHCG, and mRFM were more plantarflexed than those of DFM and OFM in the whole cycle (Fig. 4). In addition, there were significant differences in overall SPM curves for FF kinematics between the models in all planes. With respect to the point of peak angle, each model was significantly different in FF^sag and FF^trans, and some MFMs had standard deviations greater than the mean in HF^trans and FF^cor (Fig. 5).