WASICA: An effective wavelet-shrinkage based ICA model for brain fMRI data analysis

Area of Science:

• Neuroimaging
• Signal Processing
• Computational Neuroscience

Background:

• Independent Component Analysis (ICA) algorithms like Infomax and FastICA benefit from super-Gaussian properties in brain fMRI source separation.
• Enhancing super-Gaussian features through sparse approximation may improve ICA performance.

Purpose of the Study:

• To present a novel Wavelet-Shrinkage based ICA (WASICA) model for single-subject fMRI analysis.
• To improve sparse approximation and ICA-based decomposition/reconstruction for enhanced source separation.

Main Methods:

• Developed WASICA, an extension of SACICA, featuring enhanced sparse approximation coefficient set formation via wavelet-shrinkage and automatic node selection.
• Employed ICA for temporal dynamics extraction, Wavelet Packet (WP) reconstruction, and least-square-based functional networks reconstruction.

Main Results:

• Wavelet-shrinkage and automatic node selection enhanced super-Gaussian distribution in sources and mixtures.
• WASICA demonstrated superior spatial-temporal performance, source recovery, and spatial robustness compared to FastICA, Infomax, and SACICA in simulations and hybrid data.

Conclusions:

• WASICA is a promising model for brain signal separation.
• The model exhibits excellent spatial-temporal performance for functional neuroimaging analysis.