Correcting for Superficial Bias in 7T Gradient Echo fMRI

Area of Science:

• Neuroimaging
• Functional Magnetic Resonance Imaging (fMRI)
• Human Brain Anatomy

Background:

• Submillimeter ultra-high field fMRI allows for detailed cortical layer activation analysis.
• Gradient echo (GE) fMRI's blood oxygenation level dependent (BOLD) signal is biased towards superficial cortical layers, complicating laminar investigations.

Purpose of the Study:

• To compare existing and novel methods for correcting superficial bias in 7 Tesla (7T) fMRI data.
• To evaluate the effectiveness of these correction methods in computational simulations and a pilot human dataset.

Main Methods:

• Computational simulations using 7T fMRI data from a visual attention paradigm.
• Application of univariate and multivariate bias correction methods, including ROI ratio and Deming regression.
• Testing on a pilot dataset of human 7T fMRI data.

Main Results:

• Simulations identified the ROI ratio and Deming regression as the most robust methods for correcting superficial bias.
• Deming regression offers an advantage by not requiring differences in mean activation across voxels within an ROI.
• Application to pilot data showed differing layer profiles with various attention metrics, but Deming regression and ROI ratio obscured these differences, suggesting successful bias correction.

Conclusions:

• Accurate correction of superficial bias is essential to prevent erroneous conclusions in laminar analyses of GE fMRI data.
• The findings are supported by both simulation results and pilot human 7T fMRI data analysis.
• Deming regression and ROI ratio methods provide robust bias correction for laminar fMRI studies.