_

CAVEAT on using the python wrapper (see bottom)

These examples contain mainly two different views on how to use OpenMEEG with scripting languages

It concerns files compute_leadfields.* view_leadfields.* mesh.py example_inverse_problem.py view_head_model.py. The viewer works with mayavi.

It concerns files:

corticalmapping*.py tdcs.py data/canonical/* data/canonical_real/*

and uses Python tools for viewing:

om_display.py, om_compare.py, om_basics.py

The viewer works with VTK.

Here are more details:

Demo scripts to compute leadfields with OpenMEEG:

• Supports EEG, MEG, EIT and Internal potential leadfields

This folder contains a sample realistic dataset for EEG, MEG, EIT and internal potential forward modeling.

The head model is a 3 layers model with 3 nested meshes:

brain.tri, skull.tri and head.tri

To run the computation you can use the scripts:

compute_leadfields.bat

./compute_leadfields.sh

python compute_leadfields.py

The leadfields computed are stored in (matlab format):

eeg_leadfield.mat (for EEG)

meg_leadfield.mat (for MEG)

eit_leadfield.mat (for EIT)

ip_leadfield.mat (for Internal Potential)

The files used during the BEM computation are stored in the "tmp" folder.

See sample_output.txt to see what the scripts output should look like.

On a recent workstation the computation takes about:

744.66 s  344.12 s   390.41 s om_assemble -HM

• 431.74 s 188.5 s 246.5 s om_inverser
• 2689.71 s 545.1 s 719.13 s om_assemble -DSM
• 0.13 s 0.02 s 0.04 s om_assemble -H2EM
• 80.92 s 3.19 s 16.42 s om_gain -EEG
• 3.39 s. 0.88 s 1.23 s om_assemble -H2MM
• 2.33 s 0.21 s 0.95 s om_assemble -DS2MM

• 84.84 s 4.17 s 17.65 s om_gain -MEG

  173.49 s om_assemble -EITSM

  1.15 s om_gain -EEG

• canonical is a model generated remeshing the MNI template from SPM. It uses conductivities, 1., 0.0125, 1. and has a few defined source terms (8) with intensities roughly at the same scale as the model. 128 electrodes.
• canonical_real is the same model with real potentials and sensors (64) locations.

For cortical mapping:

$ python corticalmapping.py canonical_real

For TDCS:

$ python tdcs.py canonical

Beware that a temporary object has its memory released. So do not work with data provided from an OpenMEEG temporary object.

For example, having a symmetric matrix defined as:

>>> M = om.SymMatrix(100)

Taking as a numpy array the sub-matrix of M might lead to corrupted memory:

>>> mySubMat = om.asarray(M.submat(0,10,0,10))

since submat returns a newly created object that is hold by nothing, thus destroyed afterward.

Instead do keep an object pointing the newly created submatrix, and then access the numpy array form of it:

>>> subM = M.submat(0,10,0,10) >>> mySubMat = om.asarray(subM)

If you meet some difficulties running this example please contact:

openmeeg-info@lists.gforge.inria.fr

The OpenMEEG developers.