NeuroWell — EEG + AI for Early Well-being Screening (Pilot)

What it is. A research prototype that uses short EEG recordings + quick self-reports to estimate stress/mental-wellbeing risk for students. It is not a medical device.

Install

conda create -n neurowell python=3.10 -y
conda activate neurowell
pip install -r requirements.txt

Configure device

Edit config.yaml:

• device.type: muse_ble or openbci_serial
• Provide muse_mac or serial_port
• Adjust channel names to your montage.

Collect data

python src/stream_record.py

This saves data/raw/eeg_<task>.csv for tasks: baseline, nback, stroop, breathing.

Labeling

Create labels/labels.csv like:

file,label
eeg_baseline.csv,0
eeg_nback.csv,1
eeg_stroop.csv,1
eeg_breathing.csv,0

Train

python src/train_sklearn.py

Outputs:

• artifacts/model_logreg.pkl
• data/processed/features.csv
• CV ROC-AUC printed in console.

App

streamlit run app/app.py

Upload features.csv → get a risk score.

Ethics

• Obtain consent/assent; students can opt out.
• Pseudonymize data; store locally & encrypted.
• Provide school resources/helplines.