Overview

This repository contain the code needed to reproduce the results described in the paper on MEG sensor array optimization by Zhdanov, Nurminen, Iivanainen and Taulu expected to be published somewhen around 2023.

The code is written in Python and uses Anaconda for installation.

Instructions for use

The instructions assume that you have Anaconda installed.

Installation

• Create the conda environment. From the root folder of this repository (the one that contains environment.yml) run conda env create
• Activate the environment by conda activate megsim
• Install the package by pip install -e . from the root folder

Running

• Run the optimizations for the 2D and 3D imaging volumes. From the opt subfolder run python opt_run_2D.py /path/to/output/folder (for the 2D imaging volume) or python opt_run_3D.py /path/to/output/folder (for the 3D imaging volume). The optimization process will run for several days and save the results to the /path/to/output/folder.
• Reproduce the paper's figures by running the scripts from the make_figures subfolder. The scripts receive the path to the optimization results as a command-line parameter. For example, if you have your optimization results in the /path/to/output/folder, you can view the sensor geometry progression by running python sens_geom_opt.py /path/to/output/folder

Note, that you don't need to wait for the optimization process to finish to start looking at the results. The optimization saves intermediate results to the output folder as it progresses; the scripts from the make_figures subfolder can read and visualize partial results.

For more information see the comments in the files.

Notes

• The plots that require human anatomical data get the data from the MNE sample data set. If the dataset is not present on the computer, MNE functions will try to download it automatically. The location to which dataset is downloaded can be controlled by MNE_DATASETS_SAMPLE_PATH environment variable.

• For spherical coordinates, the code uses the convention of Taulu and Kajola 2005, Hill 1954, etc. That is, the angles are (theta, phi), where:

  • theta: polar (colatitudinal) coordinate; must be between 0 and pi.
  • phi: azimuthal (longitudinal) coordinate; must be between 0 and 2*pi.
    This is different from the scipy convention.