A unifying model of concurrent spatial and temporal modularity in muscle activity

Area of Science:

• Neuroscience
• Motor Control
• Computational Biology

Background:

• Movement control is thought to emerge from modular organization in the central nervous system (CNS).
• Previous studies identified muscle activation modules but used diverse decomposition methods.
• A unified framework is needed to reconcile these different approaches to muscle activation modularity.

Purpose of the Study:

• To introduce a unified framework for analyzing muscle activation modularity.
• To present a novel space-by-time decomposition model and a corresponding algorithm (sNM3F).
• To demonstrate the model's effectiveness in representing muscle activation patterns for arm pointing movements.

Main Methods:

• Developed a space-by-time decomposition to factorize muscle activations into spatial and temporal modules.
• Created a sample-based nonnegative matrix trifactorization (sNM3F) algorithm for module inference.
• Applied the method to electromyographic data from arm pointing movements.

Main Results:

• The space-by-time decomposition provides a low-dimensional, accurate, flexible, and task-relevant representation of muscle patterns.
• Extracted modules possess clear functional meaning and balance pattern replication with task discriminability.
• Results are compatible with existing models like synchronous synergies and temporal primitives, and generalize time-varying synergies.

Conclusions:

• Simultaneous, separate condensation of spatial and temporal dimensions of muscle patterns is effective.
• The space-by-time decomposition offers a unified view of hierarchical mapping from task parameters to muscle activations.
• This framework can serve as a reference for studying compositional motor control.