eeg-connectivity

EEG functional connectivity computation processes cleaned epochs across conditions and frequency bands to generate connectivity matrices using metrics like wPLI, PLV, coherence, imaginary coherence, Granger causality, or PAC. It requires frozen analysis plans and existing epoch-stage inputs.

Source-space EEG connectivity analysis is supported with leakage and orthogonalization guardrails if a source-stage exists. It applies validated metrics and issues explicit warnings about sensor-space volume conduction to ensure defensible synchronization claims.

Graph metrics calculation from EEG connectivity networks optionally thresholds connectivity matrices and outputs graph-theoretic summaries plus group aggregation artifacts. These outputs are suitable for downstream statistics and paper-ready figure generation.

EEG connectivity computation requires a frozen ANALYSIS_PLAN.md, existing epochs-stage inputs, ENVIRONMENT.json backend resolution, and consistent channel or source mappings before writing connectivity-stage outputs and updating FINDINGS.

wPLI is validated for sensor-space EEG connectivity and mitigates volume conduction artifacts better than PLV. The tool performs metric validity checks and provides explicit warnings about sensor-space synchronization limitations.

Sensor-space EEG connectivity analysis risks misleading synchronization claims due to volume conduction. The tool issues explicit warnings about these limitations and recommends source-space analysis with leakage correction when defensible connectivity claims are needed.