A new fMRI quality metric using multi-echo information: Theory, validation and implications

Area of Science:

• Neuroimaging
• Functional Magnetic Resonance Imaging (fMRI)
• Biophysics

Background:

• High-quality fMRI data is crucial for reliable results, but is often compromised by confounding factors and noise.
• Existing quality assurance (QA) metrics for fMRI do not fully utilize the information available in multi-echo (ME) acquisitions.
• There is a need for novel QA metrics that specifically leverage the unique properties of ME-fMRI data.

Purpose of the Study:

• To introduce and validate a new QA metric, termed p_BOLD, for ME-fMRI.
• To quantify the extent to which ME-fMRI signal fluctuations are dominated by Blood Oxygenation Level-Dependent (BOLD) contrast.
• To assess the utility of p_BOLD in evaluating fMRI preprocessing pipelines and its relationship with other QA metrics like TSNR.

Main Methods:

• Theoretical principles of the p_BOLD metric were established.
• p_BOLD efficacy was validated on a small dataset (N=7) with controlled BOLD fluctuation levels.
• p_BOLD was applied to a large public ME-fMRI dataset (N=439) to compare preprocessing pipelines and its relationship with TSNR.

Main Results:

• ME-based denoising strategies improved both p_BOLD and TSNR compared to basic denoising.
• Including the global signal (GS) as a regressor improved TSNR but decreased p_BOLD.
• The decrease in p_BOLD with GS regression suggests a reduction in neural BOLD fluctuations, not physiological ones.

Conclusions:

• p_BOLD is a valuable new QA metric for ME-fMRI, offering complementary information to TSNR.
• The metric can differentiate the impact of various denoising strategies on BOLD signal dominance.
• Higher p_BOLD values correlate with improved predictive power of phenotypes from functional connectivity data, highlighting its utility in downstream analyses.