Muscle synergies in upper limb amputees shoulder movements

Modular motor control is a widely accepted explanation of how the central nervous system organizes control signals to manage the redundancy of the human neuromuscoloskeletal system. For this reason, models built using this approach can be included in the myoelectric control pipeline to improve both the reliability and robustness of the latter. To do this, the first step to be solved involves a fast and accurate identification of the muscle synergies, which can be realized through non-negative matrix factorization (NNMF) of multi-muscle electromyography (EMG) recordings. In this work, the identification of muscle synergies through NNMF was investigated by exploiting simple shoulder movements in healthy and transhumeral amputee subjects, aiming to a definition of a common coordination structure between the two groups. Results suggest that the synergies identified from both groups are equivalent in the description of the multi-muscle EMG activity. Indeed, the variance accounted for (VAF) metric showed values greater than 90% for both healthy and pathological conditions, and in synergies reconstruction using a common 3-synergies vector basis. This yields promising results for the definition of faster protocols for the training of synergy-based control algorithms, shifting the problem from the one of training a subject-specific model to the one of adapting a general pattern to a subject-specific behaviour.