Multi-Band EEG Power Variability and Coherence with BFV2

Compact MATLAB analysis of EEG band-power variability and its low-frequency coherence with cerebral blood flow velocity (BFV2).

The repository studies whether slow fluctuations in EEG band power show frequency-domain association with slow fluctuations in BFV2 within a single physiological recording. It is structured as a reproducible exploratory analysis rather than a large software framework.

Project overview

The dataset includes:

• F4C4: frontal EEG channel
• P3O1, P4O2: posterior EEG channels
• BFV2: cerebral blood flow velocity
• Fs_EEG = 512 Hz
• Fs_BFV = 0.5 Hz

The analysis converts EEG into one band-power value per 2-second epoch, producing slow power-variability signals that can be compared with BFV2 on the same effective time scale.

Bands analyzed:

• delta: 0.5-4 Hz
• theta: 4-8 Hz
• alpha: 8-13 Hz
• beta: 13-30 Hz

Analysis pipeline

1. EEG channels are segmented into non-overlapping 2-second epochs.
2. Welch PSD is estimated within each epoch.
3. Band power is obtained by integrating the PSD over each frequency band.
4. Epoch-wise band power values are assembled into slow power-variability time series.
5. Magnitude-squared coherence is computed between each band-power variability signal and BFV2.
6. Coherence is summarized over 0.02-0.15 Hz using peak coherence, peak frequency, mean coherence, and area under the coherence curve.

Control analysis

Two channel-selection strategies are exported:

• primary: delta, theta, and beta from F4C4; alpha from a posterior channel when available (P3O1 preferred, then P4O2)
• same_channel_control: all bands from F4C4

The control analysis is included to make band comparisons easier to interpret when alpha is estimated from a different channel than the other bands.

Repository structure

.
|-- data/
|   `-- dataset09.mat
|-- figures/
|   |-- fig01_example_channels.png
|   |-- fig02_band_power_variability.png
|   |-- fig03_coherence_with_bfv2.png
|   `-- fig04_strategy_comparison.png
|-- results/
|   |-- analysis_outputs.mat
|   |-- strategy_comparison.csv
|   |-- summary_primary.csv
|   `-- summary_same_channel_control.csv
|-- src/
|   |-- analyzeStrategy.m
|   |-- computeBandPowerVariability.m
|   |-- defaultConfig.m
|   |-- epochSignal.m
|   |-- exportProjectOutputs.m
|   `-- loadDataset.m
|-- LICENSE
|-- main.m
`-- README.md

How to run

Open MATLAB in the repository root and run:

main

The script loads data/dataset09.mat, runs the full analysis, saves figures to figures/, and writes tabular and MAT outputs to results/.

Tested locally with MATLAB R2024b.

Outputs

The repository exports:

• example channel plots
• band-power variability plots
• coherence plots for the primary strategy
• a strategy-comparison figure
• CSV summaries for the primary and control analyses
• a MAT file containing the full exported analysis state

Scientific limits

• This is a single-recording exploratory analysis, not a population study.
• Coherence measures frequency-domain association and should not be interpreted as causality.
• The primary strategy mixes channel location with band selection, so it should be interpreted as a band-plus-channel comparison.
• The same-channel control improves interpretability but is not a formal null model or significance test.

Possible extensions

• Add a simple surrogate or permutation-based null analysis for coherence.
• Compare additional same-channel and cross-channel strategies.
• Apply the pipeline to multiple recordings if comparable datasets are available.