[BUG, MRG] fix coordinate frame in fNIRS bug

[alexrockhill]

Fixes #9627.

[alexrockhill]

Look okay to you @rob-luke?

[rob-luke]

Does this require that a user converts their montage to head frame in advance? Or is this done in the upstream function before this internal function is called?

I'm imagining the use case where your optode positions have been specified in MRI frame. How should a user load that montage?

Perhaps we should make it an option for users to specify the frame? Or if not, we should make it clear in the documentation that you must use head frame.

But I'm still learning about coordinatede frames. And I would defer to @larsoner and @agramfort opinions here.

[alexrockhill]

Does this require that a user converts their montage to head frame in advance? Or is this done in the upstream function before this internal function is called?

I'm imagining the use case where your optode positions have been specified in MRI frame. How should a user load that montage?

Perhaps we should make it an option for users to specify the frame? Or if not, we should make it clear in the documentation that you must use head frame.

But I'm still learning about coordinatede frames. And I would defer to @larsoner and @agramfort opinions here.

They are converted to head in mne.channels.montage._set_montage.

Line 800: mnt_head = _get_montage_in_head(montage)
Then later passed in Line 966: _set_montage_fnirs(info, mnt_head)

[rob-luke]

Thanks for clarifying. The fNIRS IO page isn't rendering with the example usage of set montage, but i assume it looks identical to before?

[alexrockhill]

Yes, should be identical

[larsoner]

Thanks @alexrockhill