def test_add_ragged_data_to_dynamic_table(units_table, spike_times):
write_nwb.add_ragged_data_to_dynamic_table(table=units_table,
data=spike_times,
column_name='spike_times')
assert np.allclose([1, 2, 3, 4, 5, 6], units_table['spike_times'][0])
assert np.allclose([], units_table['spike_times'][1])
assert np.allclose([13, 4, 12], units_table['spike_times'][2])
waveform_duration=duration,
cluster_depths=cluster_depths[i],
sampling_rate=30000.0)
# Add spike amps and depths
amps = {}
depths = {}
for c in cluster_info.keys():
amps[c] = spike_amps[cluster_info[c]]
depths[c] = spike_depths[cluster_info[c]]
add_ragged_data_to_dynamic_table(
table=nwb_file.units,
data=amps,
column_name='spike_amps',
column_description=
'The peak-to-trough amplitude, obtained from the template and '
'template-scaling amplitude returned by Kilosort (not from the raw data).')
add_ragged_data_to_dynamic_table(
table=nwb_file.units,
data=depths,
column_name='spike_depths',
column_description=
'The position of the center of mass of the spike on the probe, '
'determined from the principal component features returned by Kilosort. '
'The deepest channel on the probe is depth=0, and the most superficial is depth=3820.'
)
################################################################################
# WRITE TO FILE
def neural_data(nwb_file):
"""
Add in probes as devices and electrode groups.
"""
probe_descriptions = pd.read_csv('probes.description.tsv', sep='\t')
probe_descriptions = list(probe_descriptions['description'])
electrode_groups = list()
for i in range(len(probe_descriptions)):
probe_device = Device(name=str(i))
probe_electrode_group = ElectrodeGroup(
name='Probe' + str(i + 1),
description='Neuropixels Phase3A opt3',
device=probe_device,
location='')
nwb_file.add_device(probe_device)
electrode_groups.append(probe_electrode_group)
nwb_file.add_electrode_group(probe_electrode_group)
# CHANNELS
"""
Add channel information into the Electrode Table.
"""
# Read data
insertion_df = pd.read_csv('probes.insertion.tsv', sep='\t')
insertion_df['group_name'] = insertion_df.index.values
channel_site = read_npy_file('channels.site.npy')
channel_brain = pd.read_csv('channels.brainLocation.tsv', sep='\t')
channel_probes = read_npy_file('channels.probe.npy')
channel_probes = np.ravel(channel_probes.astype(int))
channel_table = pd.DataFrame(columns=['group_name'])
channel_table['group_name'] = channel_probes
channel_table = channel_table.merge(insertion_df, 'left', 'group_name')
entry_point_rl = np.array(channel_table['entry_point_rl'])
entry_point_ap = np.array(channel_table['entry_point_ap'])
axial_angle = np.array(channel_table['axial_angle'])
vertical_angle = np.array(channel_table['vertical_angle'])
horizontal_angle = np.array(channel_table['horizontal_angle'])
distance_advanced = np.array(channel_table['distance_advanced'])
locations = np.array(channel_brain['allen_ontology'])
groups = np.asarray([electrode_groups[c] for c in channel_probes])
channel_site_pos = read_npy_file('channels.sitePositions.npy')
for i in range(len(groups)):
nwb_file.add_electrode(x=float('NaN'),
y=float('NaN'),
z=float('NaN'),
imp=float('NaN'),
location=str(locations[i]),
group=groups[i],
filtering='none')
# Add Electrode columns
nwb_file.add_electrode_column(
name='site_id',
description='The site number, in within-probe numbering, of the channel '
'(in practice for this dataset this always starts at zero and '
'counts up to 383 on each probe so is equivalent to the channel '
'number - but if switches had been used, the site number could '
'have been different than the channel number).',
data=np.ravel(channel_site))
nwb_file.add_electrode_column(
name='site_position',
description=
'The x- and y-position of the site relative to the face of the probe '
'(where the first column is across the face of the probe laterally '
'and the second is the position along the length of the probe; '
'the sites nearest the tip have second column=0).',
data=channel_site_pos)
nwb_file.add_electrode_column(
name='ccf_ap',
description=
'The AP position in Allen Institute\'s Common Coordinate Framework.',
data=np.array(channel_brain['ccf_ap']))
nwb_file.add_electrode_column(
name='ccf_dv',
description=
'The DV position in Allen Institute\'s Common Coordinate Framework.',
data=np.array(channel_brain['ccf_dv']))
nwb_file.add_electrode_column(
name='ccf_lr',
description=
'The LR position in Allen Institute\'s Common Coordinate Framework.',
data=np.array(channel_brain['ccf_lr']))
# Insertion
nwb_file.add_electrode_column(
name='entry_point_rl',
description=
'mediolateral position of probe entry point relative to midline (microns). '
'Positive means right',
data=entry_point_rl)
nwb_file.add_electrode_column(
name='entry_point_ap',
description=
'anteroposterior position of probe entry point relative to bregma (microns). '
'Positive means anterior',
data=entry_point_ap)
nwb_file.add_electrode_column(
name='vertical_angle',
description='vertical angle of probe (degrees). Zero means horizontal. '
'Positive means pointing down',
data=vertical_angle)
nwb_file.add_electrode_column(
name='horizontal_angle',
description=
'horizontal angle of probe (degrees), after vertical rotation. '
'Zero means anterior. Positive means counterclockwise (i.e. left).',
data=horizontal_angle)
nwb_file.add_electrode_column(
name='axial_angle',
description=
'axial angle of probe (degrees). Zero means that without vertical and horizontal rotations, '
'the probe contacts would be pointing up. Positive means "counterclockwise.',
data=axial_angle)
nwb_file.add_electrode_column(
name='distance_advanced',
description=
'How far the probe was moved forward from its entry point. (microns).',
data=distance_advanced)
# CLUSTERS & SPIKES
"""
Add cluster information into the Unit Table.
"""
# Read data
cluster_probe = read_npy_file('clusters.probes.npy')
cluster_probe = np.ravel(cluster_probe.astype(int))
cluster_channel = read_npy_file('clusters.peakChannel.npy')
cluster_depths = read_npy_file('clusters.depths.npy')
phy_annotations = np.ravel(read_npy_file('clusters._phy_annotation.npy'))
waveform_chans = read_npy_file('clusters.templateWaveformChans.npy')
waveform_chans = waveform_chans.astype(int)
waveform = read_npy_file('clusters.templateWaveforms.npy')
waveform_duration = read_npy_file('clusters.waveformDuration.npy')
spike_to_clusters = read_npy_file('spikes.clusters.npy')
spike_times = read_npy_file('spikes.times.npy')
spike_amps = read_npy_file('spikes.amps.npy')
spike_depths = read_npy_file('spikes.depths.npy')
# Sorting spikes into clusters
cluster_info = dict()
for i in range(len(spike_to_clusters)):
s = int(spike_to_clusters[i])
if s not in cluster_info:
cluster_info[s] = [i]
else:
cluster_info[s].append(i)
# Add Unit Columns
nwb_file.add_unit_column(
name='peak_channel',
description=
'The channel number of the location of the peak of the cluster\'s waveform.'
)
nwb_file.add_unit_column(
name='waveform_duration',
description=
'The trough-to-peak duration of the waveform on the peak channel.')
nwb_file.add_unit_column(
name='phy_annotations',
description=
'0 = noise (these are already excluded and don\'t appear in this dataset '
'at all); 1 = MUA (i.e. presumed to contain spikes from multiple neurons; '
'these are not analyzed in any analyses in the paper); 2 = Good (manually '
'labeled); 3 = Unsorted. In this dataset \'Good\' was applied in a few but '
'not all datasets to included neurons, so in general the neurons with '
'_phy_annotation>=2 are the ones that should be included.',
)
nwb_file.add_unit_column(
name='cluster_depths',
description=
'The position of the center of mass of the template of the cluster, '
'relative to the probe. The deepest channel on the probe is depth=0, '
'and the most superficial is depth=3820. Units: µm',
)
nwb_file.add_unit_column(
name='sampling_rate',
description='Sampling rate, in Hz.',
)
# Add Units by cluster
for i in cluster_info:
c = cluster_info[i]
times = np.array(spike_times[c])
annotations = phy_annotations[i]
annotations = annotations.astype(int)
channel = cluster_channel[i]
channel = channel.astype(int)
duration = waveform_duration[i]
duration = duration.astype(int)
nwb_file.add_unit(spike_times=np.ravel(times),
electrodes=waveform_chans[i, :],
electrode_group=electrode_groups[cluster_probe[i]],
waveform_mean=waveform[i, :, :],
id=i,
phy_annotations=annotations,
peak_channel=channel,
waveform_duration=duration,
cluster_depths=cluster_depths[i],
sampling_rate=30000.0)
# Add spike amps and depths
amps = {}
depths = {}
for c in cluster_info.keys():
amps[c] = spike_amps[cluster_info[c]]
depths[c] = spike_depths[cluster_info[c]]
add_ragged_data_to_dynamic_table(
table=nwb_file.units,
data=amps,
column_name='spike_amps',
column_description=
'The peak-to-trough amplitude, obtained from the template and '
'template-scaling amplitude returned by Kilosort (not from the raw data).'
)
add_ragged_data_to_dynamic_table(
table=nwb_file.units,
data=depths,
column_name='spike_depths',
column_description=
'The position of the center of mass of the spike on the probe, '
'determined from the principal component features returned by Kilosort. '
'The deepest channel on the probe is depth=0, and the most superficial is depth=3820.'
)