Page Comparison

...

This app aims at computing the mean transformation matrix across all runs of a MEG session to correct the head position of all runs thanks to this matrix. This mean transformation matrix is stored in an empty .fif file, which then can be passed as a parameter in mne.preprocessing.maxwell_filter (destination parameter).

GitHub repository

...

• For input: neuro/meg/fif

• For output: neuro/meg/fif

Inputs of the App:

...

2. The MEG signal we want to preprocess

To be used in a pipeline, this App must also take the MEG signal in .fif that will be preprocessed (it must be one of the MEG files given in the first input). This is important because all the optional files will be loaded with this input and then will be used in the next Apps.

Output of the App:

Besides, in the output directory, there are the optional files we will use in a next App and the MEG file we want to preprocess. This is mandatory, because the App that will be run after app-mean-transformation-matrix will take as input all the files in app-mean-transformation-matrix’s output directory corresponding to the datatype of the input of the next App .

Steps:

The python file mean-transformation-matrix.py is composed of several functions:

• mean_transformation_matrix()

1. Extract the transformation matrix from each file

2. Create info object of the empty .fif file

3. Compute the mean of all matrices across all files and add it to the info of the empty .fif file

4. Create fake data to put into the .fif file

5. Create raw object

• main()

1. Read all the input files by parsing the config.json

2. Raise a ValueError if only one file was given

3. Read the MEG data we want to preprocess later and save it

4. Read the optional files that will be used in the next preprocessing Apps and, if they exist, copy them in the output directory

5. Apply mean_transformation_matrix()

(see How to run Apps one after another).

Parameters of the App

The parameters of the App correspond to the parameters passed to the Python functions used in the App. They are listed in the config.json.example when you test your App locally, then when you register you App on Brainlife, you enter them and a config.json is created by Brainlife.

Here, we used the parameters set by default by mne.

Next improvements

• Rename it into “app-compute-realigment-matrix”?

• Allow the user to compute a destination.fif that is not the mean of all the runs but a specific run? To do so, it would be best not to save the entire .fif file (i.e. the MEG signal) but to extract the raw.info[“dev_head_t"]["trans"] from this file and put it in an empty fif file, like we do when we compute the mean of the runs (it would save space).