Extraction of Time-Varying Spatiotemporal Networks Using Parameter-Tuned Constrained IVA

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Systems Neuroscience

Background:

• Dynamic functional connectivity (dFC) analysis reveals changing brain network associations during scanning.
• Existing dFC methods often assume stationary spatial networks, limiting their ability to capture full spatiotemporal dynamics.
• A method is needed to model variability in both spatial and temporal domains for more comprehensive network extraction.

Purpose of the Study:

• To develop and validate a two-stage method for extracting time-varying spatial and temporal features from fMRI data.
• To address limitations of Independent Vector Analysis (IVA) with a large number of datasets.
• To identify functionally relevant brain connections that differ between healthy controls and schizophrenia patients.

Main Methods:

• A two-stage approach utilizing group-independent component analysis (GICA) and a parameter-tuned constrained Independent Vector Analysis (IVA) framework.
• GICA is used to extract reference signals, followed by IVA to estimate time-varying representations while preserving variability.
• The method was applied to large-scale resting-state fMRI data from healthy controls and schizophrenia patients.

Main Results:

• The proposed method effectively captures temporal variability in brain networks from resting-state fMRI data.
• It successfully mitigates performance issues associated with IVA when handling a large number of datasets.
• The approach identified significantly different functional connections between healthy controls and schizophrenia patients, outperforming GICA alone.

Conclusions:

• The developed two-stage IVA-based method enhances the extraction of dynamic spatiotemporal brain networks.
• This technique offers improved sensitivity in detecting group differences in functional connectivity compared to traditional GICA.
• The findings highlight the potential of this method for understanding brain disorders like schizophrenia.