eeg-ica

What problem does it solve?

EEG analysis pipelines often suffer from eye, muscle, heartbeat, and line-noise artifacts that are slow and error-prone to remove by manual inspection.

Core Features & Use Cases

• Fits ICA per subject and transfers the solution correctly: it fits on a 1 Hz high-pass copy (to improve decomposition quality) and then applies the ICA weights to the original analysis-bandpass data.
• Auto-labels components using ICLabel and removes artifact classes: it predicts component types (eye blink, eye movement, muscle, heart, line noise, channel noise) and excludes those whose confidence exceeds a threshold.
• Safeguards manual review for ambiguity: components with high ambiguity or low confidence are flagged and paused for human confirmation rather than being blindly rejected.

Quick Start

Run ICA artifact removal for your study by executing: /eeg-ica projects/my-study --method infomax --n_components 0.99 --threshold 0.7