Cost of normalization: effects of maximum-peak normalization on muscle synergies obtained from transcranial magnetic stimulation

In our daily activities, even the simplest movement requires the activation of many muscles simultaneously. Muscles synergies represent a fundamental organizing principle of the neuromuscular system, acted in the central nervous system that recruits muscles and groupings them together within units to efficiently execute motor tasks in a coordinate manner. Understanding the concept of muscle synergies and their role in motor control is crucial in various fields. Research has shown that voluntary muscle synergies are highly adaptable and can be modified based on task requirements, training, and rehabilitation. The present thesis focuses on a novel approach, which has been introduced recently that complements conventional voluntary muscle synergies (task-muscle synergies): muscle synergies obtained from transcranial magnetic stimulation using motor evoked potentials (evoked-muscle synergies). Commonly, muscle synergies are obtained through a dimensionality-reduction algorithm such as non-negative matrix factorization. A crucial step before factorization of EMG signals usually involves signal normalization allowing to make comparisons among different conditions and/or different individuals. Many task-muscle synergy studies applied normalization to maximum peak-to-peak amplitude (max-peak normalization) without posing questions such as if the normalization is necessary or could have an impact on results. It has been suggested that applying max-peak normalization could lead to changes in muscle contributions, even in muscles that do not actually show much activation, referred to as 'amplifying effects'; these effects would be more evident in signals with small amplitude. My thesis aims at investigating the effects of the application of max-peak normalization on evoked-muscle synergies. Firstly, the effects of max-peak normalization were examined on MEP amplitude, before factorization. For this, contributions of each muscle to a dataset were computed and changes of contributions of each muscle were assessed through correlation analyses. Results demonstrated that max-peak normalization changes muscle contributions significantly, which are different from non-normalized data (original). This could be interpreted as a hint that max-peak normalization razes spatial information, which is muscle, in data. Consequently, such alterations in contributions must also affect the composition of the corresponding synergy vectors, which therefore requires cautious interpretation. To assess such, a further analysis was conducted comparing similarity of synergy vectors including or excluding MEP with small amplitude similar to noise. Results demonstrated that max-peak normalization shows differences in synergy spatial structure obtained with small MEP-amplitude similar to noise, comparing with non-normalized data (original) and those with a different normalization. This suggests that the amplifying effects are present with max-peak normalization, which is shown for the first time in the field of muscle synergies. This thesis aims at enhancing understanding of the use of normalization in analyzing evoked-muscle synergies and raise awareness about the effects brought by this preprocessing step. The findings highlight an impact resulting from the application of max-peak normalization, as it can have a substantial influence on outcomes and interpretations in the field of muscle synergies.