add framework for TWA head pos as MF destination

[drammock]

Opening this for early feedback. So far it looks like it won't require a very big diff, but perhaps that's the naïvety of not having tested it yet 🙈. Could probably be tested with funloc once that's finished / merged.

Before merging …

• Changelog has been updated (docs/source/dev.md.inc)

[larsoner]

Would be great to test using funloc, FYI could also in principle use this dataset but it's probably nice to have that it continue testing as it is

mne-bids-pipeline/mne_bids_pipeline/tests/configs/config_ds004229.py

Line 18 in dd1f025

mf_destination = mne.transforms.translation( # rotate backward and move up

[drammock]

OK this is now ready for another set of eyes again. Still TODO is setting up a proper test for it; I haven't had time yet to see if there's a suitable dataset (funloc won't work without monkeypatching, as it doesn't have multiple runs). That said, it appears to be working locally.

It feels like TWA-HP should actually be a separate "step" between _02_headpos and _03_maxfilter. As implemented now, if run 1 and run 2 for a given subj/sess are happening in parallel, they may both try to compute TWA-HP, where really it should only ever need to happen once.

Right now there is no logging.

I suppose it might be possible to tack TWA-HP computation onto the end of the headpos step instead of the beginning of the maxfilter step... then in maxfilter step we just check the config and load the destination Transform... that might make both logging and parallelization issues easier... WDYT?

[larsoner]

I suppose it might be possible to tack TWA-HP computation onto the end of the headpos step instead of the beginning of the maxfilter step... then in maxfilter step we just check the config and load the destination Transform... that might make both logging and parallelization issues easier... WDYT?

Yeah if you're computing time-varying head positions, I think it makes sense to always compute the time-weighted average and save it. It's very quick to compute and the resulting file is tiny. Then whether or not you use it in the maxwell_filter step depends on whether or not you set the destination to 'twa' or whatever