μ-STAR: A novel framework for spatio-temporal M/EEG source imaging optimized by microstates.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biophysics

Background:

• Electroencephalography (EEG) and Magnetoencephalography (MEG) offer noninvasive brain activity monitoring with high spatiotemporal resolution.
• Conventional source imaging models often oversimplify by assuming stable source activity, neglecting transient brain dynamics.
• Accurate source reconstruction is crucial for understanding neural mechanisms but remains a challenging, ill-posed problem.

Purpose of the Study:

• Introduce μ-STAR, a novel source imaging method combining microstate analysis with spatio-temporal Bayesian modeling.
• Improve the reconstruction of transient brain activity dynamics by incorporating optimal temporal windowing.
• Validate the performance and robustness of μ-STAR against existing benchmark models using simulations and real-world data.

Main Methods:

• Applied microstate analysis to automatically determine optimal time window lengths for quasi-stable source activity patterns.
• Developed a spatio-temporal Bayesian model utilizing user-specific spatial priors and data-driven temporal basis functions.
• Employed variational Bayesian inference and convex analysis for computationally efficient source reconstruction.

Main Results:

• Numerical simulations demonstrated that incorporating optimal temporal window lengths significantly enhanced source reconstruction accuracy.
• μ-STAR exhibited robust performance across diverse simulation settings (e.g., varying source numbers, signal-to-noise ratios, and source depths).
• Real-data validation on EEG datasets yielded neurophysiologically plausible source activity reconstructions, consistent with known neural substrates.

Conclusions:

• μ-STAR effectively addresses the limitations of conventional methods by accounting for transient brain dynamics.
• The method provides accurate and robust source imaging, outperforming several benchmark models.
• μ-STAR shows significant potential for various neuroimaging applications, advancing the understanding of brain activity.