Characterizing neural phase-space trajectories via Principal Louvain Clustering

Area of Science:

• Neuroscience
• Computational Neuroscience
• Data Science

Background:

• Neuroscience datasets are growing in size and complexity.
• Identifying spatiotemporal patterns is crucial for understanding brain function.
• Multivariate dimension-reduction techniques are essential for analyzing high-dimensional neural data.

Purpose of the Study:

• To introduce Principal Louvain Clustering (PLC), a novel method for analyzing complex neuroscience data.
• To identify spatiotemporal patterns and clusters within low-dimensional subspaces of neural activity.
• To demonstrate the application of PLC to multisite local field potential (LFP) recordings.

Main Methods:

• Proposed Principal Louvain Clustering (PLC) method.
• Applied PLC to time-varying spectral dynamics of LFP recordings from awake behaving mice.
• Analyzed data from prefrontal cortex, hippocampus, and parietal cortex during environmental exploration.

Main Results:

• PLC identified consistent subspaces and clusters across different animals.
• The identified clusters were modulated by the animals' ongoing behavior.
• Demonstrated the method's ability to capture salient features in neural dynamics.

Conclusions:

• PLC is a valuable addition to methods for characterizing dynamics in high-dimensional datasets.
• The method uses a reduced number of parameters for efficient analysis.
• PLC is adaptable for use with other neurophysiological datasets like EEG and MEG.