EEG-BCI Deep Learning Review Pipeline

This repository contains the systematic review pipeline and results for our manuscript "Deep Learning Methods for EEG-Based Brain-Computer Interfaces: A Systematic Review". The pipeline implements a reproducible workflow for literature search, deduplication, and screening of EEG-BCI papers.

Review Results Summary

Our systematic review identified 196 papers that met all inclusion criteria. The complete PRISMA flow is:

• 1,150 records identified through database searches
  • PubMed: 1,000 records
  • CrossRef: 100 records
  • Repositories: 50 records
• 1,142 records after duplicate removal
• 740 full-text articles assessed
• 196 studies included in final synthesis

Key artifact files:

• artifacts/prisma_counts.json — Complete PRISMA flow statistics
• artifacts/included.bib — BibTeX entries for all 196 included papers
• artifacts/records_deduped.csv — Full screening decisions and metadata
• artifacts/search_results/ — Raw database search results

To reproduce the review pipeline:

1. Configure config.yaml (date range, sources). The current config uses 2019-01-01 to 2024-12-31.
2. Ensure API keys are set (see .env / README notes).
3. Run the pipeline:

python run_pipeline.py --config config.yaml

Pipeline outputs will be written to artifacts/. If you are preparing the repository for review and do not want to include raw data, replace the placeholder files in artifacts/ and paper-summary/ with the real exports and the 49-papers-complete-table.csv containing all extracted fields.

Getting Started

The pipeline is configured to search and screen papers from 2019-2024. To run your own systematic review:

1. Set up your environment:

python -m venv .venv
source .venv/bin/activate  # On Unix/macOS
# Or .venv\Scripts\activate on Windows
pip install -r requirements.txt

# Copy and configure API keys
cp .env.example .env
# Edit .env to add your API keys

2. Configure and run:

# Test configuration
python run_pipeline.py --config config.yaml --dry-run

# Run full pipeline
python run_pipeline.py --config config.yaml

The pipeline will automatically:

• Search configured databases
• Remove duplicate records
• Apply inclusion/exclusion criteria
• Generate PRISMA statistics
• Export results to BibTeX

Repository Structure

Key components:

• run_pipeline.py - Main pipeline orchestrator
• config.yaml - Pipeline configuration (date ranges, sources, etc.)
• connectors/ - Database connectors (PubMed, IEEE, CrossRef, etc.)
• screeners/ - Title/abstract and full-text screening logic
• utils/ - Helper utilities (deduplication, PDF processing, etc.)
• exporters/ - Output formatters (CSV, BibTeX, JSON)

Notes for Reproducibility

• The pipeline uses Google's Gemini API for enhanced screening. Set GEMINI_API_KEY in .env.
• Rate limits are enforced for database searches to respect terms of service.
• All screening decisions are logged with rationales in records_deduped.csv.
• The PRISMA diagram and statistics are automatically generated from the data.

Citation

If you use this pipeline or our systematic review results, please cite:

@article{medrag2025,
  title={Deep Learning Methods for EEG-Based Brain-Computer Interfaces: A Systematic Review},
  author={[Author list]},
  journal={[Journal]},
  year={2025},
  status={Under Review}
}