Instantaneous brain dynamics mapped to a continuous state space

Area of Science:

• Neuroscience
• Computational Neuroscience
• Data Science

Background:

• Functional Magnetic Resonance Imaging (fMRI) measures whole-brain activity but faces interpretation challenges due to high-dimensional data.
• Understanding the brain's dynamical operations requires methods to simplify and interpret complex neural activity patterns.

Purpose of the Study:

• To develop a novel analytical framework for interpreting high-dimensional whole-brain dynamics.
• To segment brain activity into distinct states and understand their relationship during rest and task conditions.
• To map these brain states onto the brain's surface for better visualization and understanding.

Main Methods:

• Applied scale transformations in spectral, spatial, and relational domains to fMRI data.
• Utilized a wavelet filter bank for instantaneous multispectral dynamics and Independent Component Analysis (ICA) for spatial projection.
• Embedded correlation distance over wavelet-ICA state vectors onto a lower-dimensional space to analyze state-space dynamics.

Main Results:

• Successfully segmented empirical brain activity into a continuum of stimulus-dependent brain states.
• Identified that resting brain activity includes states similar to, as well as distinct from, task-active states.
• Revealed specific patterns of brain activity supporting experimentally-defined states by back-projecting dynamical state space segments.

Conclusions:

• The developed method effectively simplifies and interprets complex, high-dimensional brain activity.
• Brain states during rest and task exhibit both overlap and distinctiveness, providing a nuanced view of brain dynamics.
• This approach offers a powerful tool for mapping and understanding the neural underpinnings of cognitive states.