negida-fe-toolbox

MATLAB toolbox for dominant frequency-based EEG feature extraction in Lewy body disease patients.

Frequency Extraction (FE) Toolbox for EEG Analysis

Overview

FE-Toolbox is a MATLAB-based toolbox for extracting the dominant-frequency EEG features, including:

• Dominant Frequency (DF)
• Dominant Frequency Variability (DFV)
• Dominant Frequency Prevalence (DFP)
• Individual Alpha Frequency (IAF)
• Absolute and relative band power

The toolbox is optimized for EEG slowing and pre-alpha analysis in neurodegenerative disorders.

Frequency Bands

The power spectrum is divided into the following bands:

• Delta: 3.0–3.5 Hz
• Theta: 4.0–5.5 Hz
• Pre-alpha: 6.0–7.5 Hz
• Alpha: 8.0–12.0 Hz

Band limits are editable in the GUI. The GUI also offers Beta (13–30 Hz) and Gamma (30–45 Hz); these lie outside the 3–14 Hz filter and are reported as NaN whenever that filter is applied.

Method

• Spectra are computed per channel and epoch with a Hamming window on a 0.5 Hz frequency grid (2-s segments). Epochs longer than 2 s are split into 2-s segments with 50% overlap and averaged (Welch), so the whole epoch is used. Shorter epochs are zero-padded to the same grid.
• DF: frequency of maximum power in each epoch, searched within the filter passband (3–14 Hz when the GUI filter is applied).
• DFV: standard deviation of DF across epochs (Overall column), and of the within-band peak frequency for each band.
• DFP: percentage of epochs whose DF falls in each band. Epochs whose DF falls in no band (for example 12.5–14 Hz) are counted in an Other column, so each row sums to 100%.
• Relative power: band power as a percentage of the total power summed over all valid bands, from the epoch-averaged spectrum.
• IAF: peak and centre-of-gravity frequency of the epoch-averaged spectrum within 7–13 Hz.
• Continuous (non-epoched) datasets are cut into non-overlapping 2-s epochs.

Usage

GUI

Run FE_Toolbox, load one or more EEGLAB .set files, choose an output folder, tick the outputs you want, and press Run. You will be asked whether to apply a 3–14 Hz band-pass filter; choose No if your data are already filtered, so they are not filtered twice.

Each ticked output is written to its own workbook with one sheet per subject. EEG_Results_Summary.xlsx (always written) lists, for every subject and channel, the number of epochs, mean DF, DFV and IAF.

Scripted

EEG = pop_loadset('S001.set');
EEG = pop_eegfiltnew(EEG, 'locutoff', 3, 'hicutoff', 14);
R = fe_extract(EEG, 'Passband', [3 14]);
R.DF          % channels x epochs
R.RelPow      % channels x bands (%)
fe_write_results(R, 'S001', outDir, struct('DF', true, 'DFV', true));

fe_extract also accepts a numeric [channels x samples x epochs] array: fe_extract(data, Fs, ...). See help fe_extract for all options.

Preprocessing

EEG data should be preprocessed before using this toolbox for feature extraction. It is recommended that EEG data are band-pass filtered at 3–14 Hz before feature extraction.

Requirements

• MATLAB R2020a or later
• EEGLAB (for loading .set files and filtering; the firfilt plugin provides pop_eegfiltnew)

Testing

addpath(pwd); results = runtests('tests/test_fe_toolbox.m')

The tests use synthetic EEG with known answers. The end-to-end GUI test runs when EEGLAB is on the path or its folder is set in the EEGLAB_PATH environment variable.

License

© 2026 Ahmed Negida
Released under CC BY-NC-ND 4.0 (see LICENSE)

Publication

This toolbox was used to extract the EEG features reported in:

Negida A, Lageman SK, Ono K, Ahsan M, Mukhopadhyay N, Barrett MJ. Resting-state EEG features of cognitive fluctuations across the Lewy body disease spectrum. Alzheimer's Research & Therapy. 2026. doi:10.1186/s13195-026-02148-8

Citation

If you use this toolbox, please cite the publication above and the software (see CITATION.cff).