Multi-Subject Analysis for Brain Developmental Patterns Discovery via Tensor Decomposition of MEG Data.

Area of Science:

• Neuroscience
• Developmental Neuroscience
• Computational Neuroscience

Background:

• Understanding brain development relies on identifying informative signatures from electrophysiological signals like magnetoencephalography (MEG).
• Existing methods often underutilize the rich, multidimensional nature of MEG data for extracting meaningful brain patterns.
• Tensor factorization offers a way to model the multidimensionality of MEG data, revealing latent brain processes.

Purpose of the Study:

• To develop and validate a tensor-based approach for extracting developmental signatures from multi-subject MEG data in pediatric cohorts.
• To leverage canonical polyadic (CP) decomposition for identifying latent spatiotemporal components in MEG data.
• To utilize these extracted components for group-level statistical inference and cognitive assessment in children.

Main Methods:

• Applied canonical polyadic (CP) decomposition to multi-subject MEG data to extract latent spatiotemporal components.
• Employed hierarchical clustering in conjunction with CP decomposition to identify distinct event-related field (ERF) components.
• Utilized extracted components for group-level statistical inference and correlation with cognitive measures.

Main Results:

• Extracted typical early and late latency event-related field (ERF) components using CP decomposition and clustering.
• Identified significant correlations between extracted MEG components and cognitive domains including attention, memory, executive function, and language.
• Demonstrated that tensor-based inference yields signatures descriptive of multidimensional MEG data and discriminative of performance groups.

Conclusions:

• Tensor-based group-level statistical inference of MEG data provides effective signatures for understanding multidimensional neural processes.
• The developed method can reveal group differences in brain patterns and cognitive function in healthy children.
• This approach offers a valuable tool for assessing child developmental status and brain function from electrophysiological measurements.