Factorization of EMG via muscle synergies in walking task: Evaluation of intra-subject and inter-subject variability

Abstract

Muscle synergies have reached great popularity in the last years, to explain how the central nervous system (CNS) reduces the dimensionality of muscle activation in the execution of a specific motor task. The aim of the present paper is the quantification of the intra-subject and inter-subject variability related to the factorization of EMG data via muscle synergies. The activity of twelve muscles of the dominant leg were recorded in six healthy subjects, who were asked to perform five repetitions of level walking task at self-selected speed. A non-negative matrix factorization was performed to extract the muscle synergies. The goodness of the EMG reconstruction was assessed via the Variability Account For (VAF) between acquired and reconstructed data, for both global data and each muscle. The reliability of these parameters was evaluated through the Intra-class Correlation Coefficient (ICC) and the intra-subject and inter-subject variability of the same parameters via the Coefficient of Variation (CoV). Three, four and five muscle resulted the number of synergies selected at least in one repetition for one subject. ICC was excellent for the global reconstruction, while from poor to excellent reliability was shown for the reconstruction of each muscle. The CoV of the global reconstruction was always lower than 1.8%, decreasing with the number of synergies. By considering each muscle individually, the intra-subject variability was rarely above the 10.0% and only for the 3-synergies model. The same trend was confirmed by computing the inter-subject variability, never above the 12.6%, 6.0 % and 4.8 % and related to mean values greater than 77.8 %, 81.9 % and 92.3 % for 3-, 4- and 5- synergies models, respectively. Consequently, 4- and 5- synergies models allow to consistently reconstruct the EMG signals across repetitions and subjects. These results strengthen the possibility to use the EMG factorization via muscle synergies to describe the behavior of the CNS in the achievement of a motor task.