feature request: bad channel exclusion from common_reference function

[WeissShahaf]

Dear devs,
I appreciate your hard work. spikeinterface is becoming a really great system for nerual data processing.

Please consider implementing a bad channel exclusion option for common_reference function.
similiar to catGT's -chnexcl={prb;chans} ;this probe, exclude listed chans from ap loccar, gblcar, gfix

maybe something like this:
si.common_reference(rec, operator="median",chnexcl=bad_channel_ids)

Thank you,
Shahaf

[alejoe91]

Hi @WeissShahaf

Thanks for the support :)

You can already do that! If you have a list of channels to exclude, you can use the remove_channels() function:

bad_channel_ids = [...]

recording_clean = recording.remove_channels(bad_channel_ids)
recording_clean_cmr = spre.common_reference(recording_clean)

Is this what you had in mind?

[WeissShahaf]

@alejoe91
no, the functionality i'm describing is to exclude channels from the CMR function, but keep them in the recording.

[alejoe91]

I see, that makes sense. Basically just an optional exclude_channel_ids argument. Should be straightforward to implement!

On the other hand, median should be quite robust against outliers, so unless you have many many channels to exclude, you should be fine.

[h-mayorquin]

Does this accomplish what you want: @WeissShahaf

from spikeinterface.core import generate_recording, aggregate_channels
from spikeinterface.preprocessing import common_reference


recording = generate_recording(num_channels=4, durations=[10], seed=0)
recording = recording.rename_channels(new_channel_ids=["a", "b", "c", "d"])
bad_channels = ["b"]
recording_for_crm = recording.remove_channels(remove_channel_ids=bad_channels)
recordin_bad_chanenls = recording.select_channels(channel_ids=bad_channels)


common_reference = common_reference(recording_for_crm, reference='global')

recording_where_some_channels_are_excluded_from_crm = aggregate_channels([common_reference, recordin_bad_chanenls])
recording_where_some_channels_are_excluded_from_crm.get_traces()

[alejoe91]

@WeissShahaf fixed here #3139, but with a slight different design:

You can optionally pass a list of channels to compute the median on with the ref_channel_ids existing argument

[h-mayorquin]

To be more precise you would do

from spikeinterface.core import generate_recording
from spikeinterface.preprocessing import common_reference


recording = generate_recording(num_channels=4, durations=[10], seed=0)
recording = recording.rename_channels(new_channel_ids=["a", "b", "c", "d"])
good_channels = ["a", "b", "c"]


common_reference = common_reference(recording_for_crm, reference='global',  ref_channel_ids=good_channels)

Is that correct, @alejoe91 ? Feel free to edit so this remain as a snipet for anyone that finds this thread.

I am closing this as soon as merge #3139 if no one disagrees.