Fig. 5: Augmented muscle afferents enable biomimetic gait under neuromodulation.
From: Continuous neural control of a bionic limb restores biomimetic gait after amputation
a, Plots showing CTL and AMI cohort averaged bionic torque–angle trajectories at maximum walking speeds with varying levels of residual muscle afferents: non-biomimetic (n = 4 CTL), low level (n = 3 CTL), moderate level (n = 3 AMI) and high level (n = 4 AMI). Corresponding agonist–antagonist muscle afferents for each group are reported (mean ± s.e.m.). For comparison, a biologically intact ankle torque–angle trajectory is shown on the right11. Note how neuromodulated bionic gait evolves toward natural walking as residual muscle afferents increase. b, Subjects’ neuromodulated bionic functional metrics at their maximum speeds are plotted against their agonist–antagonist muscle afferents. Pearson correlations (r), 95% CI and slopes (m) are reported (n = 14, **P = 0.0011, ***P < 2.3 × 10−4). For comparison, values for each biomechanical metric from a biologically intact population are indicated by dashed lines49,50,52. The torque–angle trajectories and correlation plots for other gait conditions are shown in Extended Data Figs. 6–8. c, PCA applied to bionic metrics for all gait conditions tested is used to evaluate gross bionic functionality. To have consistent positive signs for biomimetic features, the signs of negative peak power and net work for stair descent are reversed. The PC1 scores are plotted against subjects’ agonist–antagonist muscle afferents. Pearson correlations (r), 95% CI and slopes (m) are reported (n = 14 for unperturbed gait conditions, n = 10 for perturbed walking). Two-sided unpaired t-tests were used for PC1 comparisons (n = 7 per cohort for walking and terrain adaptation, n = 6 CTL and n = 4 AMI for perturbation response). ***P < 4.4 × 10−4.
