A Constrained ICA-EMD Model for Group Level fMRI Analysis

Area of Science:

• Neuroimaging
• Data Analysis
• Signal Processing

Background:

• Independent Component Analysis (ICA) is valuable for functional magnetic resonance imaging (fMRI) but struggles with group data analysis.
• Existing methods to adapt ICA for group studies have limitations.

Purpose of the Study:

• To propose a novel ICA-based workflow for extracting resting-state networks from fMRI group studies.
• To address the incompatibility of standard ICA with group data analysis.
• To eliminate inherent ambiguities in ICA through a constrained approach.

Main Methods:

• Utilizing empirical mode decomposition (EMD) to generate reference signals.
• Incorporating these EMD-generated references into a constrained version of ICA (cICA).
• Comparing the proposed workflow against a widely used group ICA approach for fMRI.

Main Results:

• Intrinsic modes extracted by EMD are suitable references for cICA.
• The approach yields typical resting-state patterns consistent across subjects.
• The processing pipeline produces comparable activity patterns across subjects transparently.

Conclusions:

• The proposed EMD-guided cICA workflow offers a user-friendly method for fMRI group studies.
• It balances high inter-subject similarity with the preservation of individual subject features.
• This approach enhances the analysis of resting-state networks in fMRI group data.