run_spikesorting_openephys.py

# -*- coding: utf-8 -*-
"""
Created on Wed Apr  5 14:02:41 2023 by Guido Meijer
"""

import os
import numpy as np
import json
from datetime import datetime
from probeinterface import Probe
import spikeinterface.extractors as se
import spikeinterface.preprocessing as spre
from spikeinterface.extractors.neuropixels_utils import get_neuropixels_sample_shifts
from spikeinterface.sorters import run_sorter


from spikeinterface.core import extract_waveforms

from spikeinterface.postprocessing import compute_spike_amplitudes, compute_correlograms
from spikeinterface.qualitymetrics import compute_quality_metrics
from spikeinterface.exporters import export_report

# Set data path
SPIKE_SORTER = 'kilosort3'  
DATA_FOLDER = 'D:\\NeuropixelData'

# Load in channel locations (only for data collected with OpenEphys GUI < 0.6)
channel_locs = np.load('C:\\Users\\Neuropixel\\Documents\\spikeinterface\\channel_locations.npy')

# Search for spikesort_me.flag
print('Looking for spikesort_me.flag..')
for root, directory, files in os.walk(DATA_FOLDER):
    if 'spikesort_me.flag' in files:
        print(f'\nFound spikesort_me.flag in {root}')
        
        # OpenEphys recording
        print('OpenEphys recording detected')
        
        # Load in recording
        split_path = root.split(os.sep)[:-5]
        oe_path = os.path.join(split_path[0] + os.sep, *split_path)
        rec = se.read_openephys(oe_path, stream_id=root[-1])
        
        # Get OpenEphys GUI version
        split_path = root.split(os.sep)[:-2]
        meta_path = os.path.join(split_path[0] + os.sep, *split_path)
        info = json.load(open(os.path.join(meta_path, 'structure.oebin'), 'r'))
        print(f'GUI version {info["GUI version"]}')
        
        # For OpenEphys GUI version < 6, we need to add the channel location information 
        if int(info['GUI version'][2]) < 6:
            
            # Add channel locations
            probe = Probe(ndim=2, si_units='um')
            probe.set_contacts(positions=channel_locs, shapes='circle', shape_params={'radius': 5})
            probe.set_device_channel_indices(np.arange(channel_locs.shape[0]))
            rec.set_probe(probe, in_place=True)
        
        # Correct for inter sample shifts
        rec = spre.highpass_filter(rec)
        rec = spre.phase_shift(rec)
        
        # Detect and interpolate bad channels
        channel_labels, all_channels = spre.detect_bad_channels(rec)
        out_channels = rec.channel_ids[channel_labels=="out"]
        noise_and_dead_channels = rec.channel_ids[np.isin(channel_labels, ("noise", "dead"))]
        rec = rec.remove_channels(out_channels)
        rec = spre.interpolate_bad_channels(rec, noise_and_dead_channels)
        rec = spre.highpass_spatial_filter(rec)
       
        # Run Kilosort3
        try:
            print(f'Starting spike sorting at {datetime.now().strftime("%H:%M")}')
            sort = run_sorter(SPIKE_SORTER, rec, output_folder=os.path.join(root, SPIKE_SORTER),
                              verbose=True, docker_image=True)
        except Exception as err:
            print(err)
            
            # Log error to disk
            logf = open(os.path.join(root, 'error_log.txt'), 'w')
            logf.write(str(err))
            logf.close()
            
            # Continue with next recording
            continue
        
        
        # Export to phy
        we = extract_waveforms(rec, sort, os.path.join(root, SPIKE_SORTER, 'sorter_output'), sparse=True)
        
        # some computations are done before to control all options
        compute_spike_amplitudes(we)
        compute_correlograms(we)
        compute_quality_metrics(we, metric_names=['snr', 'isi_violation', 'presence_ratio'])
        
        # the export process
        export_report(we, output_folder=os.path.join(root, SPIKE_SORTER, 'sorter_report'))
          
        # Delete spikesort_me.flag
        print(f'Done! At {datetime.now().strftime("%H:%M")}')
        os.remove(os.path.join(root, 'spikesort_me.flag'))