Large-scale probabilistic functional modes from resting state fMRI.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Activity Analysis

Background:

• Spontaneous, structured fluctuations in brain activity are observable via functional magnetic resonance imaging (fMRI) during rest.
• Interpreting these resting-state fMRI (rfMRI) signals remains a significant challenge in neuroscience.

Purpose of the Study:

• To introduce a novel method for identifying modes of coherent activity from rfMRI data.
• To characterize these modes using a probabilistic generative model that accounts for between-subject variability and hemodynamic responses.
• To enable the inference of distinct, correlated spatio-temporal brain activity modes.

Main Methods:

• Developed a probabilistic generative model for rfMRI data analysis.
• Utilized a variational framework for Bayesian inference on voxelwise rfMRI data.
• Constrained the model by the nature of between-subject variation and the hemodynamic response function.

Main Results:

• The method stably infers sets of brain activity modes with complex spatio-temporal interactions.
• Successfully identified spatial differences in brain activity modes between subjects.
• Demonstrated robust performance on both simulated and Human Connectome Project rfMRI data.

Conclusions:

• The proposed model provides a neuroscientifically desirable approach to characterizing correlated spatio-temporal brain activity.
• This method offers a powerful tool for analyzing individual differences in brain function from rfMRI data.
• The approach surpasses traditional spatial and temporal independent component analysis (ICA) in inferring complex brain activity patterns.