Related Experiment Video
Updated: Jun 20, 2026
Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging
Published on: June 30, 2018
Relating BOLD fMRI and neural oscillations through convolution and optimal linear weighting.
Johanna M Zumer1, Matthew J Brookes, Claire M Stevenson
1Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, University Park, Nottingham, NG7 2RD, UK. johanna.zumer@nottingham.ac.uk
This study links neural oscillations to blood-oxygen-level-dependent (BOLD) functional magnetic resonance imaging (fMRI) signals. High-frequency brain waves correlate positively with BOLD signals, while low-frequency waves correlate negatively, advancing our understanding of brain activity measurement.
Area of Science:
- Neuroimaging
- Cognitive Neuroscience
- Biophysics
Background:
- The precise relationship between neural activity and BOLD fMRI remains unclear.
- Invasive studies show BOLD correlates positively with high-frequency (30-150 Hz) and negatively with low-frequency (8-30 Hz) neural oscillations.
- Non-invasive methods are needed to explore this relationship.
Purpose of the Study:
- To investigate the correlation between neural activity, measured by magnetoencephalography (MEG), and BOLD fMRI signals.
- To compare time-frequency characteristics of neural oscillations with BOLD responses in the visual cortex.
- To assess the impact of hemodynamic response function (HRF) models on the correlation.
Main Methods:
- Used simultaneous BOLD fMRI (7 T) and MEG recordings during a visual stimulation paradigm.
- Employed a time-frequency beamformer for MEG source localization.
- Convolved MEG data with measured or canonical HRFs and deconvolved BOLD data for comparison.
Main Results:
- High-frequency bands (mid-gamma: 52-75 Hz, high-gamma: 75-98 Hz) showed a positive correlation with BOLD signals.
- Low-frequency bands (alpha: 8-12 Hz, beta: 12-25 Hz) exhibited a negative correlation with BOLD signals.
- Regression analysis incorporating all frequency bands improved BOLD prediction compared to stimulus timing alone, though no single band surpassed stimulus timing.
Conclusions:
- MEG successfully replicates findings from invasive recordings regarding time series correlations with BOLD data.
- Deconvolution of BOLD data yields neural estimates that correlate well with measured neural effects based on oscillation frequency.
- This study supports the link between specific neural oscillation frequencies and BOLD fMRI signals, validated through non-invasive techniques.
