The ankle–foot complex is an intricate network of bones, joints, muscles, tendons, ligaments, and mechanoreceptors working together to support body weight and facilitate locomotion (Abboud, 2002). Footwear significantly influences this complex harmonious system, altering foot biomechanics. Comparative studies demonstrate clear differences between barefoot and shod walking: barefoot gait typically has shorter steps, higher cadence, flatter foot placement, and increased plantarflexion at initial contact, resulting in lower vertical ground reaction forces. Conventional footwear restricts natural foot motions such as forefoot spread, medial arch compression, rearfoot eversion, and multi-segment foot mobility, highlighting footwear's active biomechanical role (Janisse and Janisse, 2008).

Athletic running shoes (RS) promote heel-strike landing patterns, provide impact cushioning, but limit natural foot mobility and arch deformation. Studies often simplify the tibiocalcaneal joint as a single unit and use limited shoe varieties (Franklin et al, 2015). Rocker-bottom shoes (RBS) reduce ankle joint motion demands and calf muscle loads but alter muscle activation patterns, with research frequently focusing on specific populations like diabetic or arthritic patients (Melotto et al., 2024). Climbing shoes (CS) enhance ankle stability and protection through rigid construction but significantly limit ankle mobility, shifting compensatory work proximally and potentially increasing impact forces (Haghighat et al., 2024). High-heeled shoes (HH) position the ankle in chronic plantarflexion, reduce ankle range of motion (ROM), concentrate forefoot pressure, and necessitate compensatory hip and knee flexion (Wang et al., 2016).

These diverse footwear effects highlight the need to study foot and ankle kinematics in realistic shod conditions rather than generalizing barefoot gait data. However, capturing joint motion within footwear presents significant technical challenges. Traditional motion capture methods using external markers suffer from soft tissue artifact, with marker movements relative to bones up to 8° (Nester et al., 2007). Researchers have attempted adaptations such as using sandals to expose landmarks (but sandals differ structurally from typical shoes), or creating shoe windows for direct skin marker placement, compromising shoe integrity and reducing heel counter stiffness (McHenry et al., 2019). Attaching markers externally on shoes leads to static alignment errors on the order of centimeters, inaccurately representing bone motion. Biplanar fluoroscopy has emerged as an effective solution, employing two X-ray cameras to record 3D skeletal kinematics at sub-millimeter precision by visualizing actual bones inside intact shoes in real time. This method eliminates errors from skin movement or shoe modifications, enabling accurate, markerless in vivo analysis of foot kinematics during walking (Koo et al., 2015).

Prior fluoroscopic studies have demonstrated high accuracy in capturing fine-grained joint behavior. These examples highlight fluoroscopy's potential to yield new insights into foot mechanics by directly measuring joint motion inside footwear with high fidelity. The present study aims to investigate talocrural and subtalar joint kinematics under different footwear conditions using high-precision biplanar fluoroscopy to quantify and compare dynamic hindfoot motions during barefoot versus shod gait, addressing limitations of prior marker-based analyses.

Methods.

The procedures used in this study were approved by the Institutional Review Board (IRB) of Seoul National University Bundang Hospital [IRB number: B-1711–432-004]. The volunteers included healthy men and women aged 18–30 years old. Individuals with internal medicine or orthopedic conditions that restricted activity, history of surgery, trauma, or physical deformities were excluded. All participants were healthy and injury-free at the time of data collection and provided written informed consent.

Fifteen healthy participants (5 males and 10 females) participated, with an average age of 22.9 years (±1. years) for the females and 24.2 years (±2.2 years) for the males. To minimize selection bias, female participants were randomly assigned to one of two subgroups: one group was tested on barefoot walking, RSs, HHs, and RBSs, while the other group was tested on barefoot walking, RSs, HHs, and CSs (Fig. 1). Male participants were tested on barefoot walking, RSs, RBSs, and CS. This was done to minimize radiation exposure and participant burden—including the exclusion of HH testing in males.

A dual-plane fluoroscopy system (KMC-1400ST; Gemss Medical, Gyeonggi-do, South Korea) was approved by the Korean Ministry of Food and Drug Safety. The X-ray tube of the system was operated at 55 kVp and 10 mA in continuous mode. Each X-ray image intensifier was equipped with a 1-megapixel CCD sensor camera (1024 × 1024 pixels, 14-bit depth). The cameras were synchronized to capture images at 100 fps with an exposure time of 0.1 ms. The position and orientation of the imaging system were simulated using a graphical three-dimensional (3D) design tool (SketchUp, Google, Mountain View, CA, USA) and adjusted to avoid interference with the walkway and pedestrian walking. The walkway was made of high-density polystyrene foam with a height of 120 cm, width of 60 cm, and length of 360 cm, and foam blocks were firmly connected for stability. The angle between the two imaging directions is 37.8°. The optimal position for obtaining dual-plane images of the right foot on the walkway is shown (Fig. 2).

Dual-plane radiographs of the right foot in the static standing posture were obtained to determine the reference positions of the bones. The participants practiced walking on the walkway for approximately 5 min until they became accustomed to barefoot walking. The participants were instructed to maintain a self-selected comfortable walking speed and step on a rectangle marked on the walkway with their right foot. Dual-plane fluoroscopic images were captured for 2 s when the foot hit the target position. The foot remained within the field of view of the system for < 1 s. After the experiment, dual-plane radiographs of a custom calibration phantom with 81 steel beads were obtained for each participant. A custom-developed calibration software was used to calculate the intrinsic and extrinsic parameters of the dual-plane imaging system. An aluminum plate with a grid of 3 mm holes was attached to the front panel of the image intensifier to obtain radiographs. The images were used to calculate the distortion of the imaging system using custom software.

All participants underwent computed tomography (CT) of the foot. Using CT data, 3D triangular mesh models of the tibia, talus, and calcaneus were obtained. The dual-plane images were imported into a 3D graphics environment using the intrinsic and extrinsic parameters of the dual-plane imaging system. The triangular mesh models of the tibia, talus, and calcaneus were imported into a graphics environment. The custom-developed software for the graphics environment was developed using MATLAB (MathWorks, Natick, MA, USA) and VTK (Kitware, Clifton Park, NY, USA). An edge-based 3D/2D registration algorithm was implemented to determine the positions in the 3D graphics space as a transformation matrix for each frame between heel strike and toe off. 3D/2D registration was also applied to dual-plane radiographs obtained during static standing to identify the reference bone positions (Fig. 3). The gait experiment was conducted on 10 frames at 10 % intervals from 0 % to 90 % of the stance phase.

The anatomical coordinate systems (ACS) of the tibia, talus, and calcaneus were determined based on the articular surface geometry during standing. A mesh was extracted from the medial and lateral boundaries of the talar dome and fitted to a cylinder to calculate the origin and z-axis of the talar ACS. The temporary x-axis was determined by the location of the second metatarsophalangeal joint and the protrusion at the lower rear of the calcaneus, as calculated from dual-plane radiographs. The y-axis was set as the cross-product of the z-axis and temporary x-axis. The x-axis was set as the cross-product of the y- and z-axes. The articular surface of the tibia slides over the talar dome. Therefore, the tibial ACS was set to have the same origin and orientation as the talar ACS while standing. The ACS of the calcaneus was determined using the contact surface of the talus-calcaneus joint. A mesh was extracted from the anterior and posterior boundaries of the posterior articular surface of the calcaneus and fitted to a cylinder to determine its origin. The direction of the axis was set to be the same as that of the talar ACS. The joint coordinate system recommended by the International Society of Biomechanics was used to quantify the ankle and subtalar joint kinematics (Fig. 3). Dorsiflexion, inversion, and adduction were considered positive values. Dual-plane fluoroscopy system included only a part of the entire stance phase. The measured stance period was normalized from 0 to 1, and the time scale was expressed as a percentage of the stance phase compared to the flexion angle of the tibia-calcaneus (hindfoot) reported in previous studies (Stebbins et al., 2006). The obtained kinematics were averaged at each time point. The kinematics of the hindfoot motion was obtained along the ankle and subtalar joints for comparison with a single functional joint. For the ankle joint, the DF-PF motion was measured. DF-PF, INV-EV, and IR-ER were measured for the subtalar and hindfoot joints.

To determine the statistical power, the ankle ROM during barefoot walking was assumed to be 20° (standard deviation [SD] 5°), and during shoed walking, it was assumed to be 15° (SD 5°). Based on an alpha error of 0.05 and a power of 0.8, a dependent sample size of 10 participants was calculated.

Data normality was confirmed using the Kolmogorov–Smirnov test. Paired t-tests were first performed between barefoot and RS in all 15 participants. For each specialty footwear condition, a one-way repeated-measures ANOVA was conducted on the subset of 10 participants who walked in barefoot, RS, and that specialty shoe (RBS, HH, or CS). When the ANOVA indicated a significant effect of shoe condition (p < 0.05), a Bonferroni-corrected post hoc paired comparison between the barefoot condition and that specific shoe condition was carried out. In all analyses, statistical significance was set at 0.05 (after adjustment for multiple comparisons). Descriptive statistics (mean ± standard deviation, as well as minimum, maximum, ROM, Phase 1,9 values) were calculated for key kinematic measures. All statistical analyses were performed using SPSS (Version 25, IBM Corp., Armonk, NY, USA).