A correlation-based method for extracting subject-specific components and artifacts from group-fMRI data.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Biomedical Engineering

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for understanding brain function.
• Group fMRI analysis requires effective methods to handle subject-specific variations and noise.
• Existing artifact removal methods may inadvertently remove true neural signals.

Purpose of the Study:

• To introduce a new correlation-based method, maxCorr, for extracting subject-specific components from group fMRI data.
• To demonstrate the effectiveness of maxCorr in reducing data variance and improving the identification of fMRI activations.
• To compare maxCorr's performance against CompCor, a standard artifact removal technique.

Main Methods:

• The maxCorr method identifies signal components that maximally correlate with one subject's data while minimally correlating with others.
• Subject-specific components identified by maxCorr are often associated with physiological noise (e.g., cardiac, respiratory) and movement.
• The study involved applying maxCorr to group fMRI datasets and comparing its performance with CompCor.

Main Results:

• Removing maxCorr-identified subject-specific components significantly reduced overall data variance.
• This reduction in variance led to improved statistical identification of true fMRI activations.
• MaxCorr demonstrated a lower likelihood than CompCor of removing actual stimulus-related activity, particularly in the absence of stimulus information.

Conclusions:

• MaxCorr offers a simple yet effective approach for extracting subject-specific components in group fMRI.
• The method is particularly advantageous for analyses involving naturalistic stimuli or brain decoding due to its independence from stimulus information.
• MaxCorr enhances fMRI analysis by reducing subject-specific variance and improving the detection of neural signals.