Research App: EEGLayoutLab
Milestone:
Web application walkthrough
Rotate a saved electrode layout, inspect the brain surface and standard-64 overlay, review evaluation metrics, and compare placement constraints.
Captured from the local research application using saved simulations on the Ernie head model. Playback timing is edited; the final design screen is a configuration preview, not a newly calculated result.
The problem
An electrode layout is more than a set of points on a head. Placement constraints affect which signals it can capture, and comparisons depend on the assumed sources, noise, and head model. EEGLayoutLab brings layout design and these numerical comparisons into one interactive research application.
What the application implements
- Design and inspect layouts. A 3D head viewer supports region constraints, electrode editing, coordinate import, layout comparison, and signal simulation. Users can reopen saved layouts and export coordinates or reports.
- Search within explicit constraints. A bounded search exchanges candidate electrodes while respecting placement constraints. It balances field-reconstruction and sensitivity measures; source-localization scores are evaluated after layout selection rather than used to claim an optimal solution.
- Connect interaction to numerical work. A React / Three.js interface and Node.js host coordinate Python computations, including SimNIBS-based forward modeling and linear minimum mean-square error (LMMSE) reconstruction of the reference electrode field.
- Reuse matching computations. Local result records retain computation identities. The application checks cached records and reuses matching completed or in-progress requests, allowing comparisons without repeating identical calculations.
Evaluation and boundaries
Current evaluations use the Ernie head template with simulated sources and specified noise models. Reconstruction and localization metrics describe performance within that model; they are not measured EEG classification accuracy.
Independent-head validation, measured human EEG or BCI-task performance, and a global-optimality guarantee have not been established. The project is a local research application, not a validated clinical device or a deployed BCI system.
Components and attribution
The project combines application development, constrained search, numerical evaluation, and reproducible result handling. SimNIBS and the Ernie example data retain their respective third-party licenses and attribution. A general license for the project’s original code has not yet been selected.