---

name: source-modeling-inverse description: "Guide MNE-Python source-space, BEM, forward, covariance, inverse, beamformer, dipole, morphing, and SourceEstimate workflows." disable-model-invocation: true metadata: disco-role: operating license: BSD 3-Clause

MNE-Python source modeling and inverse router

Use this sub-skill when a task moves from sensor-space MNE objects to anatomy, source spaces, forward solutions, inverse operators, beamformers, dipoles, labels, or source estimates.

Load order

1. For the staged anatomy → source space/BEM → forward → covariance → inverse workflow, read references/workflows.md.
2. For verified signatures and parameter decision notes, read references/api-reference.md.
3. Before running any source-modeling command, read references/prerequisites-and-data.md to check subjects_dir, transforms, BEM/source files, optional external tools, and sample/testing data assumptions.
4. If a task reports file, rank, orientation, coordinate-frame, or external binary errors, read references/troubleshooting.md.
5. To validate source-modeling inputs without doing heavy modeling, run or adapt scripts/source_inputs_check.py.

Own these requests

• Build or validate FreeSurfer-style anatomy inputs, source spaces, BEM models, BEM solutions, transforms, covariance files, forward solutions, and inverse operators.
• Choose between surface, volume, mixed, and discrete source spaces.
• Compute or use forward solutions with make_forward_solution, make_forward_dipole, sensitivity maps, and field maps.
• Build minimum-norm inverse operators and apply MNE, dSPM, sLORETA, eLORETA, or label-restricted inverses to Evoked or Epochs data.
• Build LCMV/DICS beamformers, dipole fits, mixed-norm inverse estimates, and source morphs.
• Work with SourceEstimate, vector/mixed/volume source estimates, labels, label time courses, parcellations, morphing, and source-space adjacency.

Route elsewhere

• Loading raw files, creating RawArray, or fixing channel metadata belongs to io-raw-data.
• Filtering, event extraction, Epochs, Evoked, rejection, ICA/SSP, and covariance preprocessing decisions belong to preprocessing-epochs-evoked.
• Source and brain visualization belongs to visualization-reporting after the source object exists.
• Cluster statistics, decoding, spectra/TFR, and source-level statistical tests belong to timefreq-stats-decoding-simulation.
• CLI setup commands such as mne setup_source_space and mne watershed_bem are introduced in cli-datasets-config; this sub-skill explains their data prerequisites and source-modeling context.

Operating rules

• Treat source modeling as a file-graph problem before it is a computation: identify Info/epochs/evoked, trans, src, bem, covariance, inverse or forward files, subject name, subjects_dir, coordinate frames, and channel picks before calling heavy functions.
• Do not invent anatomy paths or assume sample data is present. If a task lacks MRI/FreeSurfer outputs, offer sphere-model or EEG-without-MRI alternatives only when scientifically acceptable for the user's goal.
• Keep optional external systems explicit: FreeSurfer, OpenMEEG, MNE-C tools, PyVista/3D rendering, and MNE sample/testing datasets are optional prerequisites, not base-package guarantees.
• Align rank, projectors, bad channels, reference state, and covariance with the data that will enter the inverse. Mismatches usually invalidate source estimates even if code runs.
• Never require the original MNE-Python repository checkout at runtime; bundled references and scripts here are the operating context.