Application of independent component analysis with adaptive density model to complex-valued fMRI data

Area of Science:

• Neuroimaging
• Data Analysis
• Signal Processing

Background:

• Functional resonance magnetic imaging (fMRI) data presents challenges due to its complex nature and difficulty in modeling.
• Traditional analyses often discard phase information in fMRI data due to perceived noise.
• Independent Component Analysis (ICA) is a valuable tool for analyzing complex, real-world datasets like fMRI.

Purpose of the Study:

• To demonstrate the advantages of a complex ICA approach with flexible nonlinearity for fMRI data analysis.
• To compare adaptive complex ICA against fixed nonlinearity methods, especially under high noise conditions.
• To introduce novel procedures for analyzing and visualizing complex-valued fMRI results.

Main Methods:

• Development and application of a complex ICA algorithm with adaptive nonlinearity to fMRI data.
• Comparison of adaptive complex ICA with fixed nonlinearity ICA methods.
• Implementation of bivariate t-maps for multi-subject analysis and a complex-valued ICASSO for algorithm consistency evaluation.

Main Results:

• Complex ICA with adaptive nonlinearity outperforms fixed nonlinearity methods for fMRI data analysis.
• Improved estimation of spatial maps and task-related time courses, including enhanced phase estimation.
• Demonstrated effectiveness of the proposed visualization and consistency evaluation procedures.

Conclusions:

• Adaptive complex ICA is a more desirable approach for analyzing complex fMRI data, especially in high-noise environments.
• Matching the fMRI density model adaptively enhances analysis performance.
• The introduced procedures facilitate robust analysis and visualization of complex-valued fMRI results.