Fig 1.
A. Participants walked on an instrumented treadmill while wearing varying amounts of mass on their thighs, shanks, and feet. Electromyography (EMG), marker data, and ground reaction forces were collected. B. Experimental data were used to construct musculoskeletal simulations in OpenSim [49].
Fig 2.
Experimental data were preprocessed and passed into OpenSim to scale an anatomical model and generate kinematic and kinetic data. Muscle excitations from experimental EMG data were passed directly to the EMG-informed simulations in OpenSim Moco. We employed a two-stage EMG-informed simulation approach in which we prescribed experimental kinematics and dynamics. In both stages, we solved for adjusted muscle excitations and reserve torques that minimized a weighted sum of EMG tracking error (sum of squared errors between EMG and simulation-adjusted excitations), muscle excitation effort (sum of squared muscle excitations), and reserve torque effort (sum of squared reserve torques). Reserve torques at each joint provided additional strength in the model to enforce dynamic consistency; minimizing reserve torque effort was conceptually equivalent to minimizing net joint moment error. During Calibration, a subset of walking data (split into individual stance phases) was used to calibrate the magnitude of muscle excitations from experimental EMG such that muscle-generated moments most closely matched the observed net joint moments. During Execution, experimental excitations were adjusted using scale factors obtained during the Calibration stage. Muscle forces were computed from simulation-adjusted muscle excitations and the model state. Knee contact force was computed from a combination of muscle and resultant (intersegmental) forces.
Fig 3.
Knee contact force across loading conditions.
Knee contact force is shown as a percentage of stance for all loading conditions. On each plot, the unloaded (no mass) condition is represented as a dashed black line. The light and dark colored lines represent the low and high mass loading conditions, respectively, for each segment.
Fig 4.
Relationship between added mass and changes in peak knee contact force.
Percent change in (A) early- and (B) late-stance peak knee contact force was computed relative to the unloaded condition. The mean and standard deviation across participants is shown for each of the single-segment loading conditions. We fit a simple linear model to added thigh mass, shank mass, and foot mass using data from all eight loading conditions. The correlation (Pearson’s r) between observed and model-predicted outcomes was 0.71 for the early-stance model and 0.67 for the late-stance model. The portion of this model associated with each segment is plotted as a line-of-best fit over the data (p < 0.032).
Table 1.
Summary of two linear models relating added mass at each segment (expressed as % BW per leg) to percent change (%) in early- and late-stance peak knee contact force (KCF), separately. The coefficient value, 95% confidence interval, and p-value is presented for each added mass segment.
Fig 5.
Muscle versus intersegmental contributions to changes in peak knee contact force.
Muscle force contributions (shown in red) dominated total change in peak contact force across loading conditions. Each plot column represents a different segment loading condition. Standard deviation bars are shown across participants. Top row: Changes in early-stance peak contact force. Bottom row: Changes in late-stance peak contact force.
Fig 6.
Changes in temporal gait characteristics, kinematics, and kinetics.
All changes are shown relative to the unloaded (no mass) walking conditions. (A) Stride time increased with foot loading and, to a lesser extent, thigh loading. (B) Early-stance peak knee angle became more flexed with shank and foot loading. (C) Late-stance peak knee extension angle became more flexed with thigh loading. (D) Late-stance duty factor decreased with foot loading and increased slightly with thigh loading. (E) Magnitude of early-stance peak knee extension moment increased with thigh and foot loading. (F) Magnitude of late-stance peak knee flexion moment increased with thigh and foot loading. Changes in kinetics with thigh loading were explained by changes in total mass.