A novel multi-segment foot has been developed in this study based on a modified DuPont foot model [15]. One extra marker is added to the dorsum of foot at the talar head to create a midfoot segment. There are four foot segments in total: hindfoot (calcaneus), midfoot (cuneiforms, navicular, cuboid), forefoot (five metatarsals), and hallux. Intersegmental rotation angles were measured relative to the proximal foot segment using the x–z-y Cardan angle sequence and described in all three anatomical planes (dorsi/plantar flexion, inversion/eversion, internal/external rotation). Results showed that twenty-six out of twenty-seven outcome measures demonstrated strong within-subject reliability (R > 0.7).

In order to compare the current model with other multi-segment foot models in the literature, ranges of motion (ROMs) of each foot segment in the current model are listed in Table 5 for comparison with models by Lee et al. [22], Leardini et al. [20] and Jenkyn et al. [16]. Due to different foot segmentation methods and the use of x–z-y Cardan angle convention instead of the joint coordinate system by Grood and Suntay [14], our model did show differences in ROMs compared to the other three models. Our model’s forefoot ROMs, hallux ROMs and hindfoot dorsiflexion/plantar flexion ROM are comparable to Lee et al.’s model [22], but its ROMs of internal/external rotation and inversion/eversion are smaller. This is likely because the addition of midfoot now reveals motions that are inclusively measured in Lee et al.’s definition of hindfoot. Our model’s midfoot ROMs are comparable to Leardini et al.’s model [20].

The multi-segment foot kinetic measures generally agree with Bruening et al. [4] and Saraswat et al. [24] in terms of joint patterns and magnitudes. However, the current model measured a large dorsiflexion moment, abduction moment and energy absorption (power valley) at the 1st metatarsophalangeal (MTP) joint in early stance phase that were absent in Bruening et al. [4] and Saraswat et al. [24]. It is speculated that the methods adopted to measure GRFs may have led to differences in hallux kinetic profiles. Bruening et al.’s study [4] used two adjacent force plates to partition the foot at the mid-tarsal joint and 1st MTP joint and measure GRFs applied to adjacent foot segments separately. Saraswat et al.’s study [24] combined GRF data with plantar pressure data to identify when and for how long a foot segment was loaded and used the ratio of segmental vertical forces to partition shear forces accordingly. Both methods omitted the inertial effects of foot segments during walking since the whole foot has a relatively small inertial effects compared to the whole body [4, 24]. When only one force plate is used to measure GRFs and the inertial effects of foot segments are disregarded, a calculation method “CPcross” has been proposed in the literature. It quantifies GRFs only when the center of pressure crossed anterior to the joint [6, 27]. If we apply “CPcross” to our model, GRFs sequentially crossed the proximal joint of the hindfoot, midfoot, forefoot and hallux at around 0%, 12%, 23%, and 41% of a gait cycle respectively, and thus joint moments and powers for each foot segment will be calculated and plotted starting at the corresponding time point when GRFs crossed its proximal joint. Figure 6 shows the new multi-segment foot kinetic profiles of our model using “CPcross”. It is based on Fig. 5, with plots before GRFs crossed the proximal joint of each foot segment ignored (covered by grey opaque blocks). The hallux generated a peak plantar flexion moment of 0.14Nm/kg, a peak adduction moment of 0.03Nm/kg, and the minimal power was -0.59W/kg. The new hallux (1st MTP joint) moment and power curves, in particular, are highly similar in patterns and magnitude compared to Bruening et al.’s [4] and Saraswat et al.’s [24] models.

Intersegmental joint moments and powers for foot segments in a gait cycle using the “CPcross” method

The reduced repeatability of midfoot motions (0.3 ≤ R < 0.7) is likely due to the smaller ROM of the midfoot relative to its adjacent segments, so that small errors in marker placement may have a larger effect on the calculated orientation of the segment-fixed axes. Additionally, the midfoot lacks a typical movement pattern during a gait cycle in our sample likely because it is supposed to be stable in healthy individuals during walking, especially in the frontal and transverse planes. The midfoot is worth tracking clinically since reduced or increased mobility of the midfoot during walking can suggest pathological gait when compared to healthy controls [3, 19, 24].

There were some limitations with this study. First, the masses and inertial properties for the foot segments were assumed based on anthropometric data from De Leva et al. [10] and Drillis et al. [12] rather than experimentally determined. In the absence of better data, these assumptions are considered reasonable and close to the actual properties. Small deviations likely have little effect on the kinetic output since the GRF applied to the foot is much larger in comparison to the small masses of foot segments. Second, this study used only one force plate without foot partitioning when measuring GRFs in this study. As a result, GRFs could be oversimplified when more than one segment is on the ground (e.g., during midstance). The breakdown of GRFs applied to each foot segment upon contact with the ground cannot be detected due to limited spatial and temporal resolutions of the force plate. Moreover, the direction of joint moments could be influenced by the relative position between the overall center of pressure on the force plate and the centers of mass of segments. However, the current model quantified the inertial effects of foot segments that have been neglected in previous multi-segment kinetic foot models. Only two participants were tested for test–retest reliability. Although gait patterns of healthy individuals tend to be more stable and predictable compared to individuals with pathologies, results generated from such a small sample size are prone to error and bias. Despite the small sample size, twenty-four out of twenty-seven outcome measures showed R > 0.7 for test–retest comparisons. A larger sample size is needed in future studies for test–retest reliability of our foot model.

The current multi-segment kinematic and kinetic foot model is feasible for use in a clinical setting for several reasons. The model requires only eleven markers on the foot. The markers used are individual spheres and small in size, so even on small feet there is sufficient room for placement. The model defines four foot segments, including a midfoot, so that clinically important motions within foot can be captured. The model does not require any gait-altering movement protocols or extra equipment. Moreover, the majority of kinematic and kinetic calculations are done with existing motion capture software (Cortex 7.0). This is convenient for researchers and clinicians that lack knowledge in creating special coding.

In conclusion, this study supports the within-subject reliability and validity of this novel multi-segment kinematic and kinetic foot model. This model was able to quantify intersegmental foot kinetics, including the joint moments and powers at the midfoot with the current model provides novel data to the field of foot biomechanics. It can be a clinically useful tool for research and assessments on clinical populations, help us better understand foot/ankle pathologies, and potentially inform treatments like exercise and orthoses.