EEG Processing

EEG signal-processing workflow for offline preprocessing and classical analysis of resting-state and oddball-style recordings. The repository is organized around a practical MATLAB pipeline for filtering, epoching, bad-channel screening, ICA-based artifact handling, ERP estimation, and time-frequency analysis, with saved intermediate datasets and representative figures kept alongside the analysis code.

The emphasis is methodological clarity rather than novelty claims: this is a compact end-to-end EEG methods project, not a benchmark dataset release or a general-purpose software framework. Motor-imagery decoding that previously appeared in this repository has been moved to a separate project so the scope here stays focused on preprocessing and classical EEG analysis.

Project Scope

• preprocessing from raw EEG to PreprocessStep1
• ICA-centered artifact analysis across several recording configurations
• completion of the main preprocessing branch to cleaned PreprocessStep2 datasets
• within-subject and group-level ERP analysis
• within-subject and group-level time-frequency analysis using continuous wavelet transforms
• a small appendix on PCA/ICA source-separation concepts

Workflow

1. pipeline/01_preprocessing Detrending, filtering, event alignment, epoch extraction, and bad-channel screening.

2. pipeline/02_artifact_removal_ica ICA case studies plus pipeline completion to cleaned, re-referenced PreprocessStep2 outputs.

3. pipeline/03_erp_analysis Within-subject and grand-average ERP estimation with waveform and scalp-topography summaries.

4. pipeline/04_time_frequency_analysis CWT-based power analysis with subject-level spectrograms and group-level alpha summaries.

5. appendix/ica_source_separation_concepts Supporting method-development note kept outside the core analysis path.

Outputs

Workflow stage	Main saved outputs	Analysis level
Preprocessing	sub-035_PreprocessStep1.mat, preprocessing QC figures	Single-subject
ICA and pipeline completion	sub-003_PreprocessStep2.mat, sub-035_PreprocessStep2.mat, IC inspection and cleaning figures	Single-subject
ERP analysis	within-subject ERP figures, WSA_allsubjects.mat, grand-average ERP figures	Single-subject and group-level
Time-frequency analysis	subject-level CWT figures, TF_Power_GA.mat, grand-average alpha/topography figures	Single-subject and group-level

Representative Figures

Filtering quality-control view used during preprocessing.

Example artifact-removal result from the preprocessing-completion branch.

Grand-average ERP scalp summary for the oddball branch.

Group-level alpha-band topography from the CWT workflow.

Repository Layout

EEG-Processing/
|-- pipeline/
|   |-- 01_preprocessing/
|   |-- 02_artifact_removal_ica/
|   |-- 03_erp_analysis/
|   `-- 04_time_frequency_analysis/
|-- appendix/
|   `-- ica_source_separation_concepts/
|-- scripts/
|   |-- run_preprocessing.m
|   |-- run_ica_workflows.m
|   |-- run_erp_analysis.m
|   |-- run_time_frequency_analysis.m
|   |-- run_full_pipeline.m
|   `-- validate_outputs.m
`-- README.md

Requirements

• MATLAB
• Signal Processing Toolbox
• EEGLAB for IC inspection, scalp maps, and interpolation-dependent steps

The repository is intended as a reproducible research workflow with MATLAB scripts, preserved intermediate datasets, and curated outputs. It is not packaged as a standalone EEG toolbox.

Usage

Run the workflow from the repository root in MATLAB:

addpath('scripts');
run_preprocessing();
run_ica_workflows(true);
run_erp_analysis();
run_time_frequency_analysis();

Or run the full sequence:

addpath('scripts');
run_full_pipeline(true);

Check for expected artifacts:

addpath('scripts');
validate_outputs();

Notes

• Each workflow directory includes a local README.md describing inputs, outputs, and representative figures.
• Outputs are stored next to the scripts that generate them so preprocessing decisions remain easy to trace.
• The repository favors readability and reproducibility over heavy abstraction.

Limitations

• The oddball branch is small in subject count, so the group-level outputs should be read as a compact technical study rather than a population-scale neuroscience claim.
• ICA rejection still depends on EEGLAB-assisted inspection and manual judgment.
• The project focuses on classical EEG processing and interpretation rather than machine-learning benchmarking.

Future Improvements

• record MATLAB and EEGLAB version metadata alongside exported outputs
• add lightweight figure-regeneration checks for easier reproducibility
• expand data-provenance notes where source metadata are available