Person-specific modelling of spinal motoneurons: towards customized neurorehabilitation

Human movement results from the interaction between the nervous and musculoskeletal systems. Key elements mediating such interplay are the alpha-motoneurons (α-MNs). These α-MNs, organized in muscle-specific pools (i.e., every skeletal muscle in the human body is innervated by its own pool of α-MNs), represent the final common pathway of the central nervous system. Accordingly, a pool of α-MNs integrates all synaptic inputs (i.e., from brain, cerebellum, muscle spindles, etc.) to generate the neural drive controlling the activation of skeletal muscle fibers (i.e., motor units). The dynamics underlying the conversion of the synaptic inputs into a the neural drive are majorly dependent on the properties of the α-MN pool, which vary widely across subjects. Moreover, α-MN pool properties adapt over time due to exercise, ageing or neuronal lesions, with direct consequences for motor control. Therefore, the ability to measure in vivo the behavior and underlying properties of human α-MN pools is key for understanding person-specific neural dynamics, assessing neuro-muscular adaptations and, in case of neural injury, for developing customized motor-restoring technologies that maximize recovery after lesion. In this context, extensive research endeavors have been dedicated to measure α-MN activity from human muscles, including intra-muscular and high-density electromyography (EMG) decomposition. Nevertheless, due to intrinsic physiological and technological limitations of these approaches (e.g., spatial filtering, action potential superposition, etc.), getting access to the activity of complete pool of α-MNs in vivo remains a challenge. Even more daunting is studying the physiological properties and mechanisms underlying person-specific α-MN pool dynamics. Despite significant advances in signal processing and computational neuronal modelling, current approaches are still unable to effectively bridge this gap, partly due to the lack of connection between α-MN modelling frameworks and in vivo measurements from human subjects. Therefore, this thesis introduces a novel methodology combining high-density EMG decomposition, computational neuronal modelling and metaheuristic-based parameter optimization tools for creating person- and muscle-specific in silico models of complete α-MN pools. This personalized models, able to capture and reproduce in vivo α-MN dynamic, represent a fundamental step for gaining further insight into the physiological aspects of human motor control. The structure of this dissertation is comprised of six chapters. The first constitutes a comprehensive introduction to the fields of neuro-mechanics, computational neuroscience and motor control. Chapter two entails an in-depth analysis of the α-MN model parameters associated to in vivo human α-MN activity, which are key for defining the optimal set of parameters that must be calibrated to create in silico α-MN copies. The third chapter constitutes the first published record a in silico α-MN pool models reproducing the in vivo activity of human silico α-MNs during the execution of motor tasks. Thereby showcasing the potential of the proposed approach for enhancing HD-EMG-based estimations of the neural drive. Subsequently, chapter 4 expands this methodology through the addition of task-dependent synaptic gains, thus enabling the generalization of this models to multiple motor conditions, including varying levels of force and rate of force development. Lastly, a case study is presented where this methodology is applied to evaluate the effect that transcutaneous electrical stimulation exerts over the excitability of α-MNs in a spinal cord injury subject, which is followed by a general discussion on the major outcomes, limitations and perspectives of this person-specific modelling approach for bridging the gap in current neurorehabilitation and motor-restoring technologies.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.