add OTB (EMG) reader #13303
add OTB (EMG) reader #13303
Conversation
| from ...utils import _check_fname, fill_doc, logger, verbose, warn | ||
| from ..base import BaseRaw | ||
|
|
||
| OTB_PLUS_DTYPE = np.int16 # the file I have is int16; TBD if *all* OTB+ are like that |
There was a problem hiding this comment.
@OTBioelettronica there are several TODO items in the code here that maybe you can easily answer. Here's the first: are the data in .sig files always 16-bit signed integer?
| # TODO English-centric; any value of "sex" not starting with "m" or "f" will get | ||
| # classed as "unknown". TBD if non-English-like values in the XML are possible | ||
| # (e.g. maybe the recording GUI is localized but the XML values are not?) |
There was a problem hiding this comment.
@OTBioelettronica is "sex" always "M" or "F" in these files or can users input whatever they like in this field? If left unspecified, will the <sex> node just be missing from the XML file?
| # TODO verify these are the only non-data channel IDs | ||
| NON_DATA_CHS = ("Quaternion", "BufferChannel", "RampChannel") |
There was a problem hiding this comment.
@OTBioelettronica we need to distinguish which channels are EMG data from other channel types. Are these 3 the only non-EMG channel IDs, or are there others?
| fnames = fid.getnames() | ||
| # the .sig file is the binary channel data | ||
| sig_fname = [_fname for _fname in fnames if _fname.endswith(".sig")] | ||
| assert len(sig_fname) == 1 # TODO is this a valid assumption? |
There was a problem hiding this comment.
is there guaranteed to be only 1 .sig file in each OTB+ archive? The MATLAB code suggests "yes" but it's not totally clear.
| # TODO better to call these "emg_1" etc? should we left-zeropad ix? | ||
| ch_names.append(f"{ch_id}_{ix}" if ch_id in dupl_ids else ch_id) |
There was a problem hiding this comment.
right now we get channel names like "HD10MM0804_0", "HD10MM0804_1", etc. Seems like "EMG_000", "EMG_001" might be more user-friendly.
| # TODO verify ch_type for quats, buffer channel, and ramp channel | ||
| ch_types.append("misc" if ch_id in NON_DATA_CHS else "emg") |
There was a problem hiding this comment.
@larsoner is misc the right channel type for these?
| # TODO is the factor of 1000 (copied from the MATLAB code) to convert Volts | ||
| # milliVolts? If so we should remove it (in MNE we always store in SI units, so | ||
| # we should keep it in V) | ||
| self._data = _data * (1000 * power_supply / 2**bit_depth / gains)[:, np.newaxis] |
There was a problem hiding this comment.
@OTBioelettronica can you verify that the data in the .sig file are recorded in Volts?
I think MNE-Python is better scope-wise than MNE-BIDS. Might make sense to add a MNE-EMG (or another package?) like we have MNE-NIRS if we expect a lot of EMG-specific functionality. In that case, the reader could live here or there. I think my vote is to start this |
agreed
Yes, I think following what we have with mne-nirs would be good. |
This implements a Raw reader for OTB files (EMG data). There are 2 formats, OTB4 and OTB+; so far only OTB+ implemented. Based in part on a MATLAB reader, with permission to translate and relicense as BSD-3, granted here: OTBioelettronica/OTB-Python#2 (comment)
.matfiles with source-of-truth for the signal values from the test data files.cc @OTBioelettronica @klotz-t @pranavm19 @agramfort