FieldTrip

Sensor locations are described by the elec or grad field in the data object. These definitions of the sensors can contain fewer or more channels that present in the data, i.e., you can have bipolar EOG channels that do not have a unique position on the scalp, but you can also have reference gradiometers in the MEG system that do not have a signal attached to them.

As of September 23, 2011 we updated the description of how the sensors are defined in FieldTrip. The electrode definition contains the following fields:

elec.label   % cell-array of length N with the label of each channel
elec.chanpos % Nx3 matrix with the cartesian coordinates of each channel
elec.elecpos % Mx3 matrix with the cartesian coordinates of each electrode

Note that there is typically a one-to-one match between electrodes and channels, but in principle channels and electrodes can refer to different entities. In the context of EEG, one could take a local bipolar derivation, in which each 'channel' represents the potential difference between two electrodes, and it makes sense to define its position as the mean of the two electrodes' positions.

The old electrode definition contained the following fields:

elec.pnt   % Nx3 matrix with carthesian coordinates of each electrodes
elec.label % cell-array of length N with the label of each electrode

The gradiometer definition generally consists of multiple coils per channel, e.g. two coils for a 1st order gradiometer in which the orientation of the coils is opposite. Each coil is described separately and one large matrix (grad.tra: can be sparse) has to be given that defines how the forward computed field is combined over the coils to generate the output of each channel. The gradiometer definition consists of the following fields as of September 23, 2011:

grad.coilpos   % Mx3 matrix with the position of each coil
grad.coilori   % Mx3 matrix with the orientation of each coil
grad.tra       % NxM matrix with the weight of each coil into each channel
grad.label     % cell-array of length N with the channel label
grad.chanpos   % Nx3 matrix with the position of each channel
grad.chanori   % Nx3 matrix with the orientation of each channel. 

The channel 'orientation' is needed for synthetic gradient computation for axial gradiometer or magnetometer systems. If you don't know what it means and need to construct your own grad structure, please set it to nan(N,3).

The old gradiometer definition contained the following fields:

grad.pnt   % Mx3 matrix with the position of each coil
grad.ori   % Mx3 matrix with the orientation of each coil
grad.label
grad.tra

 

The changes are motivated by the fact that the relevant information that is needed from the grad/elec structure is different for different analysis/visualization steps:

• for displaying purposes, usually the channels are the entities of relevance. Also, in the context of finding neighbours to a given channel (for clustering, or synthetic gradient computation, or interpolation as in scalpcurrentdensity or channelrepair), the channels are the relevant entities.
• for forward and inverse modelling purposes, the sensing elements, i.e. the electrodes or coils are of relevance.

Originally, FieldTrip relied on the fact that the channel positions can be recovered from the electrode/coil positions by looking into the tra-matrix, because the tra-matrix specifies which electrode/coil contributes to which channel. However, FieldTrip supports increasingly complicated tra-matrices that for example include balancing coefficients (obtained through ft_denoise_synthetic, or ft_denoise_pca), projected-out spatial topographies (obtained through a sequence of ft_componentanalysis and ft_rejectcomponent), or synthetic planar gradients (obtained through ft_megplanar). With these increasingly complicated tra-matrices, recovery of the channel positions from the coil/electrode positions is not straightforward and sometimes impossible. We decided to make the distinction between channels on the one hand, and electrodes/coils on the other hand explicit in the code.