Multiporbe recording

[mfvd]

Hello,

I've been experimenting recording with two probes simultaneously and was wondering what could be the best option to deal with the channel layouts. Would it make sense to create a single probe object with 6 shanks (3 per probe) and sort the data by shank? At first glance this seems like a viable option.

It would be nice if someone could share their experiences with such layouts :)

Cheers,
Marcelo

[alejoe91]

Hi @mfvd

You can also create 2 probes in probeinterface, then make a probegroup and add them as separate spikeinterface groups with:

recording = recording.set_probe(probegroup, group_mode='by_probe')

This will load the two probes and assign different channel groups. Then you can use the spikeinterface.sorters.run_sorter_by_property() function to spike sort separate groups individually.

[mfvd]

Hi @alejoe91

Thanks for tour quick reply!

Just to make sure I understood correctly:
Would it still be possible to run the sorter by shank or would the group property now be the probe?

[samuelgarcia]

This group_mode have a role on how to set the group property.

By probe:

recording = recording.set_probe(probegroup, group_mode='by_probe')
print(recording.get_property('group')) # show 2 groups

By shank:

recording = recording.set_probe(probegroup, group_mode='by_shank')
print(recording.get_property('group')) # show 6 groups

Then internally, spikeinterface.sorters.run_sorter_by_property use the recording.split_by(self, property='group')

Note that you can use any property also for the splitting.

[mfvd]

Hi @samuelgarcia

Ok, that makes sense. I'll give it a try!

Thanks

[mfvd]

Hi @alejoe91 @samuelgarcia

I've tried Samuel's apporach and got the following error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Input In [7], in <cell line: 31>()
     29 probegroup = pi.ProbeGroup()
     30 probegroup.add_probe(probe)
---> 31 probegroup.add_probe(probe2)
     33 # set recording channels to new probe>RHD map.
     34 # by shank to perform sorting by shank
     35 rec_probe = rec.set_probe(probegroup, 
     36                           group_mode='by_shank')

File /opt/anaconda3/envs/si_openephys/lib/python3.8/site-packages/probeinterface/probegroup.py:22, in ProbeGroup.add_probe(self, probe)
     18 """
     19 
     20 """
     21 if len(self.probes) > 0:
---> 22     self._check_compatible(probe)
     24 self.probes.append(probe)
     25 probe._probe_group = self

File /opt/anaconda3/envs/si_openephys/lib/python3.8/site-packages/probeinterface/probegroup.py:36, in ProbeGroup._check_compatible(self, probe)
     34 # check global channel maps
     35 self.probes.append(probe)
---> 36 self.check_global_device_wiring_and_ids()
     37 self.probes = self.probes[:-1]

File /opt/anaconda3/envs/si_openephys/lib/python3.8/site-packages/probeinterface/probegroup.py:188, in ProbeGroup.check_global_device_wiring_and_ids(self)
    185 valid_chans = chans[keep]['device_channel_indices']
    187 if valid_chans.size != np.unique(valid_chans).size:
--> 188     raise ValueError('channel device index are not unique across probes')
    190 # check unique ids for != ''
    191 all_ids = self.get_global_contact_ids()

ValueError: channel device index are not unique across probes

I set the device channel indices manually for each probe. Any idea of what error I've made?

manufacturer = 'cambridgeneurotech'
probe_name = 'ASSY-37-Fb'
probe = pi.get_probe(manufacturer, probe_name)

# Match to device (RHD headstage) map. In this case it was corrected because probeinterface was wrong.
device_channel_indices = [
    29, 19, 10, 18, 28, 30, 20, 17, 21, 31, 22, 16, 23, 27, 26, 25, 24, 7, 6, 5,
    4, 8, 9, 3, 11, 2, 12, 1, 13, 0, 14, 15]
# set channel ids based on probe map from cambridge site and RHD map.
probe.set_device_channel_indices(device_channel_indices)

### set info for PROBE_2 ###
probe2 = pi.get_probe(manufacturer, probe_name)
device_channel_indices_2 = [
    17, 20, 16, 21, 31, 23, 30, 24, 28, 19, 26, 8, 27, 10, 28, 11, 29, 12, 22, 7,
    25, 0, 4, 15, 5, 14, 9, 13, 6, 1, 3, 2]

device_channel_indices_2 = [element + 32 for element in device_channel_indices_2]
# set channel ids based on probe map from cambridge site and RHD map.
probe2.set_device_channel_indices(device_channel_indices_2)
# create and populate probegroup
probegroup = pi.ProbeGroup()
probegroup.add_probe(probe)
probegroup.add_probe(probe2)

rec_probe = rec.set_probe(probegroup, 
                          group_mode='by_shank')

[alejoe91]

As the error suggests, the device channel indices are not unique. You need to shift the second set by the number of channels in the first probe ;)

[mfvd]

Stupid error on my part...
I was certain device_channel_indices_2 = [element + 32 for element in device_channel_indices_2] was doing just that and did not spot a duplicate value in my list

thanks for the help @alejoe91

[mfvd]

I'm now having the following error:

---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
Input In [16], in <cell line: 41>()
     37 probegroup.add_probe(probe2)
     39 # set recording channels to new probe>RHD map.
     40 # by shank to perform sorting by shank
---> 41 rec_probe = rec.set_probe(probe=probegroup, 
     42                           group_mode='by_shank')

File /opt/anaconda3/envs/si_openephys/lib/python3.8/site-packages/spikeinterface/core/baserecording.py:325, in BaseRecording.set_probe(self, probe, group_mode, in_place)
    321 def set_probe(self, probe, group_mode='by_probe', in_place=False):
    322     """
    323     Wrapper on top on set_probes when there one unique probe.
    324     """
--> 325     assert isinstance(probe, Probe), 'must give Probe'
    326     probegroup = ProbeGroup()
    327     probegroup.add_probe(probe)

AssertionError: must give Probe

I set the probe indices and channel ids for the second probe and the values seem to be correct. Not sure what to do next.

### set info for PROBE_2 ###
probe2 = pi.get_probe(manufacturer, probe_name)
# Match to device (RHD headstage) map. In this case it was corrected because probeinterface was wrong.
device_channel_indices_2 = [
    17, 20, 16, 21, 31, 23, 30, 24, 18, 19, 26, 8, 27, 10, 28, 11, 29, 12, 22, 7,
    25, 0, 4, 15, 5, 14, 9, 13, 6, 1, 3, 2]

device_channel_indices_2_64 = [element + 32 for element in device_channel_indices_2]

probe2.set_device_channel_indices(device_channel_indices_2_64)

# set contact ids
contact_ids_2 = np.arange(33, 65, 1)
# set contact ids to account for second probe
probe2.set_contact_ids(contact_ids_2)

# create and populate probegroup
probegroup = pi.ProbeGroup()
probegroup.add_probe(probe)
probegroup.add_probe(probe2)

rec_probe = rec.set_probe(probegroup, 
                          group_mode='by_shank')

[samuelgarcia]

You need to do

rec_probe = rec.set_probegroup(probegroup, 
                          group_mode='by_shank')

[mfvd]

@samuelgarcia
Thanks!