Reducing the Effects of Motion Artifacts in fMRI: A Structured Matrix Completion Approach

Area of Science:

• Neuroimaging
• Biomedical Engineering
• Data Science

Background:

• Functional MRI (fMRI) is crucial for brain research but susceptible to motion artifacts.
• Existing motion correction methods are insufficient, leading to biased functional connectivity.
• Censoring corrupted data creates signal discontinuities, impacting analysis.

Purpose of the Study:

• To develop a novel method for artifact reduction in fMRI using structured low rank matrix completion.
• To recover censored fMRI data, mitigating motion-induced biases and signal discontinuities.
• To improve the accuracy of functional connectivity analysis.

Main Methods:

• Formulated artifact reduction as a super-resolved matrix recovery problem.
• Applied a low rank prior to a structured matrix derived from fMRI time series.
• Utilized a variable splitting strategy for efficient computation.
• Validated with simulations, diverse motion conditions, and ABCD study data.

Main Results:

• Reconstructed fMRI time series showed reduced errors in pairwise correlations compared to standard methods.
• Improved delineation of the default mode network in seed-based correlation analyses.
• Demonstrated effective reduction of motion-related adverse effects in fMRI.

Conclusions:

• The proposed structured low rank matrix completion method effectively mitigates motion artifacts in fMRI.
• This technique enhances the reliability and accuracy of functional connectivity and brain network analyses.
• The method offers a promising solution for improving fMRI data quality and interpretability.