External Email - Use Caution
reposting here from issue tracker for broader discussion.
explaining some context behind the KIT digitization process at NYU based on
a discussion with @massich.
so for the KIT system, the use of the labels elp and hpi had come from the
naming used in the old mne-c manual.
HPI
Head Position Indicator (HPI) points come from the head localization
procedure in the MEG with the attached coils. There are five coils for the
KIT system, these points have a specific order, and that this allows for
the coregistration of the head into MEG device space for the `dev_head_t`
computation. This is stored in either a `.mrk` file or a `.sqd` file, which
just contains this head localization procedure.
ELP
The electrode points file has been constructed to emulate the `.elp` file
structure that was produced from the old ISOTRAK pen IIRC. In this file,
there is a section for the fiducials which exists at the top, and there is
a section for the corresponding HPI points but collected in the digitizer
native space.
The NYU lab with the KIT system has been accustomed to include all 8 points
in an exported txt file from the FASKTRAK digitizer. This has been done as
a matter of lab protocol convention.
I think that ideally there should be three files as opposed to two files:
one for the HPI points (5 points), one for those points in digitizer space
(5 points), and one file for the fiducials (3 points). I don't think
including the fiducials in the points file is a good idea anymore because
this requires an arbitrary assignment of points to references. From my
discussion with @massich, it is true that someone could place the points in
any manner they sought fit but there is no way for the software to
intelligently know this.
File Format
FASKTRAK allows for exports to txt files but there isn't anything
inherently special about it, it just arbitrary number of point x 3. These
points are referenced to a reference box therefore all the digitizer data
must be re-reference to neuromag head space based on fiducials. The
polhemus file reader is just one that ignores the header and just reads the
data in as an array.
Naming
As for the names of HPI and ELP, there has been a proposal to rename the
points:
HPI --> `hpi_dev`
ELP --> `hpi`
After thinking about this, I think that HPI should remain being `hpi` and
ELP should be the one renamed to something like `hpi_isotrak` or
`hpi_isotrak_space`
HPI --> `hpi`
ELP --> `hpi_isotrak`
hope this provides some context and recommendations.
Best,
Teon