The INTERACT team established novel methodologies for creating digital copies of a person’s neuromuscular system. The first part of the project focused on modelling the behavior of a type of neural cells located in the spinal cord that are responsible for the generation of movement, i.e. the α-motor neurons. The team described how α-motor neurons generate the electrical activity controlling the activation of skeletal muscles.
NEURONAL MODELLING:
The functional structures of the α-motor neurons (i.e. soma, dendrites) were modelled as biophysical compartments capturing the anatomy and function of the neuron (e.g. soma size, channel dynamics, etc.). A major challenge was the identification of such properties from large sets of α- motor neurons in human subjects and in a non-invasive way (i.e. these studies are typically done in animal preparations using invasive techniques). To address this challenge, the team developed new non-invasive and clinically-viable system identification methods to decode the activity of α-motor neurons from multiple high-density electromyography recordings. Together with advanced techniques for signal processing, this technology allowed the identification of individual α-motor neurons during isometric and dynamic tasks such as walking. This new approach allowed creating digital versions of a person’s α-motor neurons (e.g. in silico neuronal copies) that behaved similarly to their biological "counterparts”.
The process involved two major steps. First, the calibration of in silico neuron’s parameters for reproducing the firing behavior of the experimentally-decoded α-motor neuron. Secondly, the creation of a person-specific representation of the entire α-motor neuron pool based on statistical distributions of the calibrated model parameters. These results are now leading to new concepts of human-machine interfacing that enables continuous neural of robotic of robotic leg exoskeletons and bionic legs by individuals with neuromuscular impairment.
SKELETAL MUSCLE MODELLING:
The INTERACT team also created digital copies of skeletal muscles. A new model was proposed for the estimation of musculoskeletal stiffness during dynamic movements, with a focus on the human leg. State of the art approaches largely rely on biological joint perturbation techniques, which alter musculoskeletal function and prevent measuring stiffness in natural (unperturbed) conditions. INTERACT created a new approach that is not based on joint-perturbation, but rather decodes stiffness from muscle electromyography recordings and leg kinematic data. This methodology, was recently validated against robotic joint perturbations and ultrasonography, providing direct evidence that the INTERACT approach decodes accurate stiffness estimates at multiple anatomical scales including: joint, single muscle and single tendon level for the ankle musculoskeletal complex. This is now opening to new views into how spinal neurons controls muscle impedance and resulting body motions in natural conditions with large implications for neurorehabilitation and robotic control.
Moreover, the INTERACT team employed an innovative in vitro approach to study how skeletal muscles alter their biological structure when exposed to mechanical stimuli over large periods of time i.e. four weeks. We grew muscle tissues on a chip starting from a gel-cell mix containing human induced pluripotent stem cells. We locally tuned the mechanical and chemical environment, with high precision, to influence muscle tissue formation and adaptation to mechanical stimuli across 30 days. Results showed this in vitro approach allows observing muscle adaptation with high levels of precision not attainable in vivo, laying the basis for self-adaptive muscle models.
NEURAL CONTROL OF WEARABLE ROBOTS:
The INTERACT neuro-muscular modelling technology was employed for the control of wearable robotic exoskeletons. The team conducted a series of studies that proved a crucial concept, i.e. whether neurologically injured patients could regain control of their paretic legs using exoskeletons controlled via the patient’s digital twin. Key results showed it was possible to create digital twins for both spinal cord injury and post-stroke individuals. Digital twins were used to establish a new patient-exoskeleton interface, which enabled patients to gain volitional control of exoskeletons and move again their paretic legs in ways that were not otherwise possible. More recently, the INTERACT technology has been successfully demonstrated for the continuous neural control of bionic legs, during repertoire of locomotion tasks, in individuals with transtibials with transtibial amputation.
