Continuous neural control of a bionic limb restores biomimetic gait after amputation

For centuries, the goal has been to create artificial leg replacements that match the versatility of natural limbs. Biological gait requires complex neural interplay, making neuroprosthetic emulation challenging. Current bionic legs rely on predefined robotic control architectures (finite state machines and pattern recognition) that model legged motion into discrete states based on gait phase and terrain. These systems lack continuous neuromodulation from the user. Neuroprosthetic legs fully driven by the human nervous system could unlock bionic capabilities approaching those of intact limbs. This requires high-bandwidth neuromodulation for adaptive foot positioning, shock absorption, and propulsion across diverse terrains. However, residual muscle control in standard amputations is inconsistent and involves unintended coactivation, hindering biomimetic gait even on level ground. Current bionic legs use neural inputs as auxiliary signals within intrinsic gait controllers, limiting neural control over gait detection, pattern recognition, and joint movements. Standard amputation procedures result in the loss of essential locomotor peripheral afferents, further complicating neural control. Electrical nerve stimulation has been proposed, but locomotion's reliance on reflexive spinal neural circuitry and the magnitude of afferents needed for high-bandwidth motor control remain unclear. Previous studies using electrical nerve stimulation improved gait but haven't demonstrated biomimetic gait under full nervous system control. The ability of individuals with intact legs to quickly compensate for locomotory disturbances suggests that augmenting residual afferent signaling might allow amputees to fine-tune their neural circuitry for biomimetic neuroprosthetic gait. This study hypothesized that augmenting muscle afferents within the amputated residuum, using a modified amputation technique (agonist-antagonist myoneural interface, AMI), would enable enhanced biomimetic gait adaptation in a neuroprosthetic leg with continuous gait neuromodulation.

Existing literature highlights the limitations of current bionic legs, which primarily rely on predefined robotic control architectures and lack continuous neuromodulation from the user's nervous system. These limitations stem from challenges in accurately estimating gait information and modeling human motor intent into a finite number of states. While some studies have explored the use of electrical nerve stimulation to provide afferent feedback and improve gait function, they haven't achieved fully biomimetic gait controlled by the nervous system. The complexity of legged neuromechanics, involving coordinated afferent and efferent signals, and the loss of distal tissue during standard amputations contribute to these challenges. Previous research using passive and intrinsically controlled bionic legs with electrical nerve stimulation showed improved gait function but not biomimetic gait under full neural control. The study aims to address these gaps by investigating the impact of muscle afferent augmentation on biomimetic gait neuromodulation.

This study was part of a clinical trial (NCT03913273) evaluating the effectiveness of the AMI procedure. The AMI surgically connects residual agonist-antagonist muscle pairs to mimic intact muscle dynamics, leveraging native sensory organs to generate biomimetic afferents. Seven subjects with unilateral below-knee AMI amputations and seven matched control (CTL) subjects with non-AMI amputations participated. Matching criteria included age, time since amputation, height, and weight. The neuroprosthetic interface included a powered prosthetic ankle, an electromyography (EMG) sensor unit, and flexible surface electrodes. A neuroprosthetic controller enabled continuous ankle torque neuromodulation (dorsiflexion and plantar flexion) throughout the gait cycle using EMG signals from the tibialis anterior (TA) and gastrocnemius (GAS) muscles. Importantly, the controller did not use predefined robotic algorithms. Muscle afferents were estimated using a computational model based on muscle fascicle strains and EMG signals. Gait testing included level-ground walking at multiple speeds, adaptations to slopes and stairs, and obstacle crossing. Statistical analyses involved paired and unpaired t-tests, Mann-Whitney U-tests, and Pearson correlations to compare AMI and CTL cohorts and assess the relationship between muscle afferent augmentation and biomimetic gait.

The AMI augmented agonist-antagonist muscle afferents by 18% of biologically intact values. AMI subjects showed a 41% higher maximum walking speed than CTL subjects, achieving speeds comparable to those of individuals with intact legs. AMI subjects neuromodulated distinct push-off and foot clearance ankle angles, mimicking natural joint kinematics. Increased walking speed in AMI subjects led to increased bionic ankle peak power and net work, unlike CTL subjects. AMI subjects demonstrated biomimetic slope and stair adaptations, maintaining level-ground walking mechanics while modulating shock absorption and propulsion. CTL subjects lacked these adaptations. AMI subjects neuromodulated higher peak power for adaptive propulsion and shock absorption (163-202% higher than CTL subjects). In obstacle-crossing trials, AMI subjects increased dorsiflexion in bionic swing kinematics and generated propulsive power during gait recovery, unlike CTL subjects. AMI subjects traversed the perturbation 54% faster than CTL subjects. Analysis of torque-angle trajectories at different muscle afferent levels showed the emergence of natural walking mechanics as afferents increased. A strong correlation was found between afferent augmentation and bionic ankle peak power, net work, LEK symmetry, maximum walking speed, and biomimetic terrain adaptations. Even modest afferent augmentation restored biomimetic locomotion and responses to perturbations.

This study demonstrated that augmented residual muscle-tendon afferents enable direct neuromodulation of biomimetic locomotion in transtibial amputees, achieving adaptations to varying speeds, terrains, and perturbations—a level of versatility not seen in contemporary bionic legs without intrinsic control frameworks. Muscle-tendon sensory organs deliver proprioceptive afferents, and the surgically reconstructed agonist-antagonist muscles emulate natural dynamics, generating these afferents. The capacity of augmented afferents to enable biomimetic gait, even with reduced magnitude compared to intact limbs, highlights human sensorimotor adaptability. The findings provide design principles for future neuroprosthetic interfaces by framing limb reconstruction as a neuroprosthetic design optimization problem. The results suggest that partial afferent restoration may suffice for achieving biomimetic neuroprosthetic function under full neuromodulation.

This study demonstrates that augmenting residual muscle afferents via the AMI procedure allows for continuous neural control of a bionic limb, restoring biomimetic gait in transtibial amputees. Even a modest increase in afferent signaling significantly enhances gait performance across various speeds, terrains, and perturbations. These findings provide valuable insights for the design of future neuroprosthetic interfaces and highlight the potential for achieving near-natural limb function through a combination of surgical reconstruction and advanced control algorithms. Future research should explore the long-term clinical benefits, patient experiences, and the application of these techniques to other amputation levels and wearable devices.

The study used a relatively small sample size, limiting the generalizability of the findings. The obstacle-crossing trial was conducted with a subset of participants due to scheduling constraints. The study primarily focused on muscle-tendon afferents, and future research could investigate the potential benefits of adding cutaneous feedback or exploring other surgical interventions like osseointegration. The study's scope was limited to a specific set of locomotory conditions, and future research could expand the range of tasks, including more dynamic perturbations.