NeuroConv Overwriting Units Table

[EthanKato]

Hello,

I am working on a conversion pipeline for multiprobe neuropixel data. I am using the neuroconv NWBconverter combined with built in data interfaces. When I try to write multiple streams of kilosort data (from each probe), instead of adding to the unit table and specifying the source probe, it overwrites the table with new kilosort data from another probe. Is there another method of converting kilosort data from multiple probes that conserves all the units and specifies which units came from which probe? I tried to follow the documentation on the NeuroConv "How to Link Sorted Data to Electrodes" page, but this documentation is out of date and the listed import doesn't even work (from neuroconv import SortedRecordingConverter doesn't work, instead from neuroconv.datainterfaces.ecephys.sortedrecordinginterface import SortedRecordingConverter). Perhaps I haven't looked hard enough, but are there any examples / guidelines for structuring NWB data for multiple neuropixel probes?

Thank you

[oruebel]

This sounds like an issue in NeuroConv. I filed catalystneuro/neuroconv#1447 to make sure the NeuroConv team can respond.

ChatGPT suggests merging the spike sortings using SpikeInterface, but I am not sure whether that this is the recommended approach with NeuroConv.

import spikeinterface.extractors as se
from spikeinterface.extractors import concatenate_sortings
from neuroconv.datainterfaces.ecephys.spikeinterface.spikeinterfacesortinginterface import SpikeInterfaceSortingInterface
from neuroconv import NWBConverter

# ----- Step 1: Load and rename sortings -----
# Replace with your actual Kilosort output paths
sorting1 = se.KiloSortSortingExtractor('probe1/')
sorting2 = se.KiloSortSortingExtractor('probe2/')

# Rename units to ensure uniqueness
unit_ids_1 = sorting1.get_unit_ids()
unit_ids_2 = sorting2.get_unit_ids()
sorting1_renamed = sorting1.rename_units({u: f'probe1_{u}' for u in unit_ids_1})
sorting2_renamed = sorting2.rename_units({u: f'probe2_{u}' for u in unit_ids_2})

# ----- Step 2: Combine sortings -----
combined_sorting = concatenate_sortings([sorting1_renamed, sorting2_renamed])

# ----- Step 3: Annotate units with probe name -----
probe_names = []
for unit_id in combined_sorting.get_unit_ids():
    if str(unit_id).startswith('probe1_'):
        probe_names.append('probe1')
    else:
        probe_names.append('probe2')
combined_sorting.set_property('probe_name', probe_names)

[EthanKato]

Ah! Thank you for opening up that discussion. That's probably a more suitable place for it.