MRG: Store NIRx location data in meters

[rob-luke]

What does this implement/fix?

Stores the NIRx sensor location data in meters (as specified for 3d locations here)

Details

• A test recording was completed as described: Add NIRx files recorded with NIRStar 15.2 with short channels and triggers mne-testing-data#51
• In step 4 of this test recording a figure was taken that shows the location of the detectors in the vendor software (in millimeters)
• This PR tests that the values read by MNE match the values in the vendor software (but converted to meters)

Note

The values displayed in the vendor software are in mm. However, the values stored with the recordings are in cm (I have confirmed this with the company technical support). And MNE expects m.

[codecov]

Codecov Report

Merging #6856 into master will increase coverage by 1.33%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #6856      +/-   ##
==========================================
+ Coverage   88.28%   89.61%   +1.33%     
==========================================
  Files         426      425       -1     
  Lines       76034    76018      -16     
  Branches    12402    12405       +3     
==========================================
+ Hits        67123    68125    +1002     
+ Misses       6135     5116    -1019     
- Partials     2776     2777       +1

[larsoner]

@rob-luke this code:

import mne
data_path = mne.datasets.testing.data_path()
subjects_dir = mne.datasets.sample.data_path() + '/subjects'
raw = mne.io.read_raw_nirx(data_path + '/NIRx/nirx_15_2_recording_w_short')
mne.viz.plot_alignment(
    raw.info, show_axes=True, subject='fsaverage', trans='fsaverage',
    subjects_dir=subjects_dir, verbose=True)

Now produces:

You can see that the points are on the head (good) and that the head and MRI coordinate frames are correct (good). There are a few points that are quite far away to the subject's right, though -- is this correct/expected?

[rob-luke]

Thanks @larsoner! I loaded a bunch of other montages and they all look great.

I investigated that point that is slightly off the head by looking at other montages I have with more channels in that same area. All the other points sit nicely on the head. The one that is off the head looks to be over a recessed section near the ear. I have tried to capture this in the image below (look at the red dot on the left of the image ). It could be that because the channel is defined as the half way point between a source and detector, that it can then float in space if their is a recess in the head.

So I think this looks great and will change the title to MRG. I am looking forward to uploading some more interesting montages to make testing things like this easier.

[agramfort]

@rob-luke you need to fix merge conflict

[rob-luke]

I have uploaded another test dataset here.: mne-tools/mne-testing-data#52

And running @larsoner code above on this new data I get this image, which matches well with the vendor images

[larsoner]

Thanks @rob-luke, feel free to add tests with the new dataset (as necessary) in the next PR