This is an old revision of the document!


Introduction

This page contains questions that users could ask when they analyze their data with FieldTrip. If it is possible the answers are also provided. If any documentation already exist on the FieldTrip wiki which gives an answer, the answer should point to those pages. In some cases, may we just want to point to relevant literature.

Questions and Answers


What kind of source-reconstruction methods are implemented in FieldTrip?


The Inverse source parameter estimates from EEG/MEG data page describes under the second point which are the supported methods.

  1. dipole fitting
    1. simultaneous optimisation of position, orientation and strength
    2. symmetry constrains and/or fixed position, with free orientation and strength
  2. dipole scanning
    1. dynamic imaging of coherent sources (DICS)
    2. linear constrained minimum variance (LCMV)
    3. partial canonical coherence (PCC)
    4. multiple signal classification (MUSIC)
    5. scanning for residual variance
  3. distributed source modeling
    1. minimum norm estimation with and without noise regularisation (MNE)

The reference of the ft_sourceanalysis function refers to the following methods:

cfg.method     = 'lcmv'    linear constrained minimum variance beamformer
                     'sam'     synthetic aperture magnetometry
                     'dics'    dynamic imaging of coherent sources
                     'pcc'     partial cannonical correlation/coherence
                     'mne'     minimum norm estimation
                     'loreta'  minimum norm estimation with smoothness constraint
                     'rv'      scan residual variance with single dipole
                     'music'   multiple signal classification
                     'mvl'   multivariate Laplace source localization
                     


What is the difference between the methods?

What kind of source-reconstruction method should I use?

  • Does it depend on the data? (EEG vs. MEG, oscillations vs event-related, realistic vs. non-realistic headmodel)


Event-related Field/Potential + time-course: MNE
Oscillatory activity + at certain point in time: beamforming (dics)
beamforming: lcmv - ?
(see Hesse, Jensen (2010) and Background of the MNE tutorial)

  • Does it depend on a priori hypothesis of the source involved? (cortical sheets vs. 3D grid)
  • Does it depend on what kind of information I am interested in? (e.g. changes in time or not)

What kind of data I need for source-reconstruction?


functional data, anatomical data, channel/electrode positions

Why should I use source-reconstruction?


point to introductionary literature

How should I do source-reconstruction?


depends on the specific method; available documentation in FT at the moment:

dipole fitting

  1. simultaneous optimisation of position, orientation and strength
  2. symmetry constrains and/or fixed position, with free orientation and strength

tutorial sites:
none.
example scripts:
Compute forward simulated data and apply a dipole fit
Fit a dipole to the tactile ERF after mechanical stimulation
Why is this fixme?
Source-reconstruction using two dipoles
This is under construction, but it is not really clear how this exactly relates to dipole fitting.

dipole scanning

  1. dynamic imaging of coherent sources (DICS)
  2. linear constrained minimum variance (LCMV)
  3. partial canonical coherence (PCC)
  4. multiple signal classification (MUSIC)
  5. scanning for residual variance

distributed source modeling

  1. minimum norm estimation with and without noise regularisation (MNE)

What kind of volume conduction models of the head are implemented in FieldTrip?

FAQ: What kind of volume conduction models are implemented?
Reference: ft_prepare_headmodel

  • The FAQ as it is now is not enough informative (not all methods are listed). The FAQ should refer to ft_prepare_headmodel.
  • But the methods listed in the reference of ft_prepare_headmodel are not consistent with method names (1) in the script of ft_prepare_headmodel and (2) it is not easy to match them to the articles in References to implemented methods.
  • The references to implemented methods can be probably extended.
  • Some references and short explanations of the methods can not be seen by the users because the individual functions (e.g. ft_headmodel_bemcp) are called from ft_prepare_headmodel and the help of ft_prepare_headmodel does not refer to the help of these lower level functions.
  • Some of the “methods” of ft_prepare_headmodel are not computations but reading functions of already made head models.


Methods:
EEG

  • name in help of ft_prepare_headmodel: asa
  • name in script of ft_prepare_headmodel: asa
  • article in References to implemented methods: none
  • lower-level function name: ft_headmodel_bem_asa
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: none but it is probably not applicable
  • reading function (reads in a certain type of volume conduction model)


  • name in help of ft_prepare_headmodel: bemcp
  • name in script of ft_prepare_headmodel: bemcp
  • article in References to implemented methods: none
  • lower-level function name: ft_headmodel_bemcp
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: none but the person's name who provided the code is mentioned


  • name in help of ft_prepare_headmodel: dipoli
  • name in script of ft_prepare_headmodel: dipoli
  • article in References to implemented methods: yes (Oostendorp T, van Oosterom A., 1991)
  • lower-level function name: ft_headmodel_bem_dipoli
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: yes


  • name in help of ft_prepare_headmodel: openmeeg
  • name in script of ft_prepare_headmodel: openmeeg
  • article in References to implemented methods: none
  • lower-level function name: ft_headmodel_bem_openmeeg
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: yes


  • name in help of ft_prepare_headmodel: concentricspheres
  • name in script of ft_prepare_headmodel: concentricspheres
  • article in References to implemented methods: yes (Cuffin, Cohen, 1979)
  • lower-level function name: ft_headmodel_concentricspheres
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: none


  • name in help of ft_prepare_headmodel: concentricspheres
  • name in script of ft_prepare_headmodel: concentricspheres
  • article in References to implemented methods: yes (Cuffin, Cohen, 1979)
  • lower-level function name: ft_headmodel_concentricspheres
  • explanation in help of the lower-level function: yes
  • reference in help of the lower-level function: none