Related Experiment Video
Updated: May 8, 2026
Multi-Tracer Studies of Brain Oxygen and Glucose Metabolism Using a Time-of-Flight Positron Emission Tomography-Computed Tomography Scanner
Published on: June 7, 2024
Optimizing RetroICor and RetroKCor corrections for multi-shot 3D FMRI acquisitions
Rob H N Tijssen1, Mark Jenkinson, Jonathan C W Brooks
1FMRIB Centre, Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK; University Medical Center Utrecht, Utrecht, The Netherlands.
Optimizing physiological noise correction in functional MRI (fMRI) is crucial for accurate results. This study found that tailored, volume-wide regressors, selected using criteria like the Bayesian Information Criterion (BIC), best balance variance reduction and prevent overfitting in 3D fMRI data.
Area of Science:
- Neuroimaging
- Biophysics
- Signal Processing
Background:
- Physiological noise significantly impacts the statistical significance of functional Magnetic Resonance Imaging (fMRI) activation detection.
- Accurate correction of physiological noise is essential for reliable fMRI data analysis.
Purpose of the Study:
- To systematically optimize physiological noise regression techniques for multi-shot 3D fMRI data.
- To compare the efficacy of image-space (RetroICor) versus k-space (RetroKCor) correction methods.
- To determine the optimal regressor set for various 3D fMRI acquisition parameters.
Main Methods:
- Simulations and experimental data were used to evaluate correction strategies.
- Investigated k-space corrections using real/imaginary versus magnitude/phase data.
- Employed the Bayesian Information Criterion (BIC) to select optimal regressor sets for diverse 3D fMRI acquisitions.
Main Results:
- K-space corrections are more robust on real/imaginary data compared to magnitude/phase data.
- K-space corrections did not outperform image-space corrections, even with improved physiological phase synchronization.
- The optimal noise correction model varied significantly across different acquisition techniques.
Conclusions:
- Tailored, volume-wide regressors, selected via criteria like BIC, are recommended for 3D fMRI.
- The optimal approach balances effective variance reduction with the avoidance of model overfitting.
- Customized noise correction strategies are necessary for different fMRI acquisition protocols.
