Hi Daniel,
I'm CCing to the list as this might be of interest to others.
On 29 Jun 2010, at 17:12, Daniel Wakeman <[log in to unmask]> wrote:
>
> One example of the naming convention inconsistencies is the following (I have not extensively explored this problem):
>
> When reading data in, Magnetometers are called MEGMAG; however when doing spm_dcm_ind_data they are called MEG. This behavior is very strange and confusing. I am not aware if this is also a problem for axial gradiometers; however, it again seems weird that all of these different sensor types get lumped together.
There is logic behind this, but it's slightly more complex. Channel type (MEGMAG, MEGGRAD etc.) is a property of a channel and it's stored in the header and will stay fixed as long as you don't change it. Modality is a property of the whole dataset and it is context-dependent because some distinctions are important to make in particular circumstances but not in others. Thus modality is determined based on available channel types and the answer depends on the kind of distinctions the user asks to make. MEGPLANAR is both a channel type and a modality but MEG is only a modality.
For instance, coregistration is done differently for MEG and for EEG, but in the MEG case it doesn't matter what kind of sensors are there planar or axial. So when asked about modality you should get either EEG or MEG. DCM, however, can only handle one sensor type at a time at the moment so it does need to distinguish between MEGPLANAR and MEG.
So if there is support for multimodal data in the future the code in spm_dcm_ind_data will have to be modified anyway so there is no need to make it more generic now. Furthermore, what happens if the user selects the ECD option for a dataset with just LFP channels? That code was that way before because it was actually clear whether to use ECD or LFP based on the data. The only problem was that MEGPLANAR was missing from the list of modalities.
Vladimir
|