Automatic independent component labeling for artifact removal in fMRI

Area of Science:

• Neuroimaging
• Signal Processing
• Machine Learning

Background:

• Blood oxygenation level dependent (BOLD) signals in functional magnetic resonance imaging (fMRI) are often obscured by significant noise.
• Sources of noise include motion artifacts and complex physiological patterns, complicating data analysis.

Purpose of the Study:

• To develop an automated method for reducing fMRI artifacts using independent component analysis (ICA).
• To improve the statistical analysis and accuracy of brain activation detection in fMRI data.

Main Methods:

• An automated ICA-based denoising technique was developed.
• A supervised classifier (global decision tree in a Neyman-Pearson framework) distinguished signal from noise components.
• A denoised fMRI time-series was reconstructed using only signal components.

Main Results:

• The automated method achieved a misclassification rate between 0.2 and 0.3.
• Denoising reduced Z-scores in white matter, indicating artifact removal.
• A similar, weaker reduction in Z-scores was observed in gray matter, suggesting improved signal quality.

Conclusions:

• Automated ICA-based denoising is a potentially valuable tool for enhancing fMRI data quality.
• This method can improve the accuracy of statistical analyses and the detection of brain activations.