A kernel machine-based fMRI physiological noise removal method.

Area of Science:

• Neuroimaging
• Biophysics
• Signal Processing

Background:

• Functional magnetic resonance imaging (fMRI) with blood oxygenation level dependent (BOLD) contrast is crucial for noninvasive brain mapping.
• Physiological noise from cardiac and respiratory cycles significantly challenges fMRI data quality, especially at higher resolutions.
• Existing noise removal methods often have limitations or cannot separate overlapping frequencies of BOLD signals and physiological noise.

Purpose of the Study:

• To develop and evaluate a novel method for effective physiological noise removal in fMRI data.
• To address the challenge of frequency overlap between BOLD activation and physiological noise.
• To improve the accuracy and reliability of fMRI studies by enhancing signal quality.

Main Methods:

• Utilized a nonlinear kernel machine technique, specifically kernel principal component analysis (KPCA).
• Employed a specifically identified kernel function within KPCA to differentiate BOLD signal from physiological noise.
• Evaluated the method on human fMRI data from task-related and resting-state experiments, comparing it with an adaptive filtering approach.

Main Results:

• The proposed kernel machine technique effectively identified and reduced physiological noise in fMRI data.
• The method demonstrated superior or comparable performance to existing adaptive filtering techniques.
• Successful differentiation of BOLD signal from frequency-overlapping physiological noise was achieved.

Conclusions:

• The kernel machine technique offers a robust and effective solution for physiological noise removal in fMRI.
• This approach enhances the potential for characterizing subtle neuronal changes and achieving higher spatial resolutions in fMRI studies.
• The method provides a valuable tool for improving the quality and interpretability of fMRI data across various experimental conditions.