def get_correlogramsview_data(exp, correlograms, clusters=[],
channel_group=0, clustering='main', wizard=None,
nclusters_max=None, ncorrbins=50, corrbin=.001):
clusters = np.array(clusters, dtype=np.int32)
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
cluster_groups_data = getattr(exp.channel_groups[channel_group].cluster_groups, clustering)
freq = exp.application_data.spikedetekt.sample_rate
cluster_colors = clusters_data.color[clusters]
cluster_colors = pandaize(cluster_colors, clusters)
# TODO: cache and optimize this
spike_clusters = getattr(exp.channel_groups[channel_group].spikes.clusters,
clustering)[:]
sizes = np.bincount(spike_clusters)
cluster_sizes = sizes[clusters]
clusters_selected0 = clusters
nclusters_max = nclusters_max or USERPREF['correlograms_max_nclusters']
# Subset of selected clusters if there are too many clusters.
if len(clusters_selected0) < nclusters_max:
clusters_selected = clusters_selected0
else:
clusters_selected = clusters_selected0[:nclusters_max]
correlograms = correlograms.submatrix(clusters_selected)
cluster_colors = select(cluster_colors, clusters_selected)
# Compute the baselines.
# corrbin = SETTINGS.get('correlograms.corrbin', CORRBIN_DEFAULT)
# ncorrbins = SETTINGS.get('correlograms.ncorrbins', NCORRBINS_DEFAULT)
duration = exp.channel_groups[channel_group].spikes.concatenated_time_samples[:][-1] - exp.channel_groups[channel_group].spikes.concatenated_time_samples[:][0]
duration /= freq
if duration == 0:
duration = 1.
baselines = get_baselines(cluster_sizes, duration, corrbin)
baselines = baselines[:nclusters_max,:nclusters_max]
data = dict(
correlograms=correlograms,
baselines=baselines,
clusters_selected=clusters_selected,
cluster_colors=cluster_colors,
ncorrbins=ncorrbins,
corrbin=corrbin,
keep_order=wizard,
)
return data
def get_correlogramsview_data(exp, correlograms, clusters=[],
channel_group=0, clustering='main', wizard=None,
nclusters_max=None, ncorrbins=50, corrbin=.001):
clusters = np.array(clusters, dtype=np.int32)
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
cluster_groups_data = getattr(exp.channel_groups[channel_group].cluster_groups, clustering)
freq = exp.application_data.spikedetekt.sample_rate
cluster_colors = clusters_data.color[clusters]
cluster_colors = pandaize(cluster_colors, clusters)
# TODO: cache and optimize this
spike_clusters = getattr(exp.channel_groups[channel_group].spikes.clusters,
clustering)[:]
sizes = np.bincount(spike_clusters)
cluster_sizes = sizes[clusters]
clusters_selected0 = clusters
nclusters_max = nclusters_max or USERPREF['correlograms_max_nclusters']
# Subset of selected clusters if there are too many clusters.
if len(clusters_selected0) < nclusters_max:
clusters_selected = clusters_selected0
else:
clusters_selected = clusters_selected0[:nclusters_max]
correlograms = correlograms.submatrix(clusters_selected)
cluster_colors = select(cluster_colors, clusters_selected)
# Compute the baselines.
# corrbin = SETTINGS.get('correlograms.corrbin', CORRBIN_DEFAULT)
# ncorrbins = SETTINGS.get('correlograms.ncorrbins', NCORRBINS_DEFAULT)
duration = exp.channel_groups[channel_group].spikes.concatenated_time_samples[:][-1] - exp.channel_groups[channel_group].spikes.concatenated_time_samples[:][0]
duration /= freq
if duration == 0:
duration = 1.
baselines = get_baselines(cluster_sizes, duration, corrbin)
baselines = baselines[:nclusters_max,:nclusters_max]
data = dict(
correlograms=correlograms,
baselines=baselines,
clusters_selected=clusters_selected,
cluster_colors=cluster_colors,
ncorrbins=ncorrbins,
corrbin=corrbin,
keep_order=wizard,
)
return data
def get_traceview_data(exp,
channel_group=0, clustering='main'):
if (len(exp.recordings) == 0) or exp.recordings[0].raw == None:
data = dict(
trace=None,
)
return data
rawdata = exp.recordings[0].raw
freq = exp.application_data.spikedetekt.sample_rate
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
clusters = sorted(clusters_data.keys())
spikes_data = exp.channel_groups[channel_group].spikes
channels = exp.channel_groups[channel_group].channel_order
spiketimes = spikes_data.time_samples
spikeclusters = getattr(spikes_data.clusters, clustering)[:]
_, nsamples, nchannels = spikes_data.waveforms_filtered.shape
freq = exp.application_data.spikedetekt.sample_rate
cluster_colors = pd.Series([clusters_data[cl].application_data.klustaviewa.color or 1
for cl in clusters], index=clusters)
fetdim = exp.application_data.spikedetekt.nfeatures_per_channel
s_before = exp.application_data.spikedetekt.extract_s_before
s_after = exp.application_data.spikedetekt.extract_s_after
if spikes_data.masks is not None:
spikemasks = np.zeros((spikes_data.masks.shape[0], rawdata.shape[1]))
spikemasks[:,channels] = spikes_data.masks[:, 0:fetdim*nchannels:fetdim]
cluster_colors = pandaize(cluster_colors, clusters)
data = dict(
freq=freq,
trace=rawdata,
spiketimes=spiketimes,
spikemasks=spikemasks,
spikeclusters=spikeclusters,
cluster_colors = cluster_colors,
s_before = s_before,
s_after = s_after
)
return data
def get_traceview_data(exp,
channel_group=0, clustering='main'):
if (len(exp.recordings) == 0) or exp.recordings[0].raw == None:
data = dict(
trace=None,
)
return data
rawdata = exp.recordings[0].raw
freq = exp.application_data.spikedetekt.sample_rate
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
clusters = sorted(clusters_data.keys())
spikes_data = exp.channel_groups[channel_group].spikes
channels = exp.channel_groups[channel_group].channel_order
spiketimes = spikes_data.time_samples
spikeclusters = getattr(spikes_data.clusters, clustering)[:]
_, nsamples, nchannels = spikes_data.waveforms_filtered.shape
freq = exp.application_data.spikedetekt.sample_rate
cluster_colors = pd.Series([clusters_data[cl].application_data.klustaviewa.color or 1
for cl in clusters], index=clusters)
fetdim = exp.application_data.spikedetekt.nfeatures_per_channel
s_before = exp.application_data.spikedetekt.extract_s_before
s_after = exp.application_data.spikedetekt.extract_s_after
if spikes_data.masks is not None:
spikemasks = np.zeros((spikes_data.masks.shape[0], rawdata.shape[1]))
spikemasks[:,channels] = spikes_data.masks[:, 0:fetdim*nchannels:fetdim]
cluster_colors = pandaize(cluster_colors, clusters)
data = dict(
freq=freq,
trace=rawdata,
spiketimes=spiketimes,
spikemasks=spikemasks,
spikeclusters=spikeclusters,
cluster_colors = cluster_colors,
s_before = s_before,
s_after = s_after
)
return data
def get_waveformview_data(exp, clusters=[], channel_group=0, clustering='main',
autozoom=None, wizard=None):
clusters = np.array(clusters)
fetdim = exp.application_data.spikedetekt.n_features_per_channel
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
spikes_data = exp.channel_groups[channel_group].spikes
channels_data = exp.channel_groups[channel_group].channels
channels = exp.channel_groups[channel_group].channel_order
spike_clusters = getattr(spikes_data.clusters, clustering)[:]
# spikes_selected = get_some_spikes_in_clusters(clusters, spike_clusters)
# cluster_colors = clusters_data.color[clusters]
# get colors from application data:
cluster_colors = pd.Series([_get_color(clusters_data, cl)
for cl in clusters], index=clusters)
# cluster_colors = pd.Series([
# next_color(cl)
# if cl in clusters_data else 1
# for cl in clusters], index=clusters)
if spikes_data.waveforms_filtered is None:
data = dict(
waveforms=None,
channels=channels,
clusters=None,
cluster_colors=None,
clusters_selected=clusters,
masks=None,
geometrical_positions=None,
autozoom=autozoom,
keep_order=wizard,
)
return data
_, nsamples, nchannels = spikes_data.waveforms_filtered.shape
# Find spikes to display and load the waveforms.
if len(clusters) > 0:
spikes_selected, waveforms = spikes_data.load_waveforms(clusters=clusters,
count=USERPREF['waveforms_nspikes_max_expected'])
else:
spikes_selected = []
# Bake the waveform data.
if len(spikes_selected) > 0:
waveforms = convert_dtype(waveforms, np.float32)
if spikes_data.masks is not None:
masks = spikes_data.masks[spikes_selected, 0:fetdim*nchannels:fetdim]
else:
masks = None
else:
waveforms = np.zeros((0, nsamples, nchannels), dtype=np.float32)
masks = np.ones((0, nchannels), dtype=np.float32)
if masks is None:
masks = np.ones((len(spikes_selected), nchannels), dtype=np.float32)
spike_clusters = spike_clusters[spikes_selected]
channel_positions = np.array([channels_data[ch].position
if channels_data[ch].position is not None
else (0., ch)
for ch in channels],
dtype=np.float32)
# Pandaize
waveforms = pandaize(waveforms, spikes_selected)
spike_clusters = pandaize(spike_clusters, spikes_selected)
masks = pandaize(masks, spikes_selected)
cluster_colors = pandaize(cluster_colors, clusters)
data = dict(
waveforms=waveforms,
channels=channels,
clusters=spike_clusters,
cluster_colors=cluster_colors,
clusters_selected=clusters,
masks=masks,
geometrical_positions=channel_positions,
autozoom=autozoom,
keep_order=wizard,
)
return data
def get_featureview_data(exp, clusters=[], channel_group=0, clustering='main',
nspikes_bg=None, autozoom=None,
alpha_selected=.75, alpha_background=.25,
normalization=None,
time_unit='second'):
clusters = np.array(clusters)
# TODO: add spikes=None and spikes_bg=None
fetdim = exp.application_data.spikedetekt.n_features_per_channel
channels = exp.channel_groups[channel_group].channel_order
clusters_data = getattr(exp.channel_groups[channel_group].clusters, clustering)
spikes_data = exp.channel_groups[channel_group].spikes
channels_data = exp.channel_groups[channel_group].channels
nchannels = len(channels_data)
spike_clusters = getattr(spikes_data.clusters, clustering)[:]
# cluster_colors = clusters_data.color[clusters]
# get colors from application data:
cluster_colors = pd.Series([_get_color(clusters_data, cl)
for cl in clusters], index=clusters)
# cluster_colors = pd.Series([
# next_color(cl)
# if cl in clusters_data else 1
# for cl in clusters], index=clusters)
if len(clusters) > 0:
# TODO: put fraction in user parameters
spikes_selected, fm = spikes_data.load_features_masks(clusters=clusters)
else:
spikes_selected = []
fm = np.zeros((0, spikes_data.features_masks.shape[1], 2),
dtype=spikes_data.features_masks.dtype)
fm = np.atleast_3d(fm)
features = fm[:, :, 0]
nextrafet = features.shape[1] - nchannels * fetdim
if fm.shape[2] > 1:
masks = fm[:, ::fetdim, 1]
else:
masks = None
nspikes = features.shape[0]
spiketimes_all = spikes_data.concatenated_time_samples[:]
spiketimes = spiketimes_all[spikes_selected]
spike_clusters = spike_clusters[spikes_selected]
freq = exp.application_data.spikedetekt.sample_rate
duration = spikes_data.concatenated_time_samples[len(spikes_data.concatenated_time_samples)-1]*1./freq
spikes_bg, features_bg = spikes_data.load_features_masks_bg()
features_bg = np.atleast_3d(features_bg)
features_bg = features_bg[:,:,0].copy()
spiketimes_bg = spiketimes_all[spikes_bg]
# Add extra feature for time is necessary.
if nextrafet == 0:
features = np.hstack((features, np.ones((features.shape[0], 1))))
features_bg = np.hstack((features_bg, np.ones((features_bg.shape[0], 1))))
nextrafet = 1
# Normalize features.
def _find_max(x):
if x.size == 0:
return 1.
return np.max(np.abs(x))
c = (normalization or (1. / _find_max(features_bg[:,:-nextrafet]))) if nspikes > 0 else 1.
features[:,:-nextrafet] *= c
features_bg[:,:-nextrafet] *= c
# Normalize extra features except time.
for i in range(features_bg.shape[1]-nextrafet-1, features_bg.shape[1]-1):
c = (1. / _find_max(features_bg[:,i])) if nspikes > 0 else 1.
features[:,i] *= c
features_bg[:,i] *= c
# Normalize time.
if features.size > 0:
features[:,-1] = spiketimes
features[:,-1] *= (1. / (duration * freq))
features[:,-1] = 2 * features[:,-1] - 1
features_bg[:,-1] = spiketimes_bg
features_bg[:,-1] *= (1. / (duration * freq))
features_bg[:,-1] = 2 * features_bg[:,-1] - 1
# Pandaize
features = pandaize(features, spikes_selected)
features_bg = pandaize(features_bg, spikes_bg)
if masks is not None:
masks = pandaize(masks, spikes_selected)
spiketimes = pandaize(spiketimes, spikes_selected)
spike_clusters = pandaize(spike_clusters, spikes_selected)
cluster_colors = pandaize(cluster_colors, clusters)
# nextrafet = features.shape[1] - fetdim * nchannels
data = dict(
features=features,
features_background=features_bg,
masks=masks,
spiketimes=spiketimes,
clusters=spike_clusters,
clusters_selected=clusters,
cluster_colors=cluster_colors,
nchannels=nchannels,
channels=channels,
fetdim=fetdim,
nextrafet=nextrafet,
freq=freq,
autozoom=autozoom,
duration=duration,
alpha_selected=alpha_selected,
alpha_background=alpha_background,
time_unit=time_unit,
)
return data
def get_waveformview_data(exp,
clusters=[],
channel_group=0,
clustering='main',
autozoom=None,
wizard=None):
clusters = np.array(clusters)
fetdim = exp.application_data.spikedetekt.n_features_per_channel
clusters_data = getattr(exp.channel_groups[channel_group].clusters,
clustering)
spikes_data = exp.channel_groups[channel_group].spikes
channels_data = exp.channel_groups[channel_group].channels
channels = exp.channel_groups[channel_group].channel_order
spike_clusters = getattr(spikes_data.clusters, clustering)[:]
# spikes_selected = get_some_spikes_in_clusters(clusters, spike_clusters)
# cluster_colors = clusters_data.color[clusters]
# get colors from application data:
cluster_colors = pd.Series(
[_get_color(clusters_data, cl) for cl in clusters], index=clusters)
# cluster_colors = pd.Series([
# next_color(cl)
# if cl in clusters_data else 1
# for cl in clusters], index=clusters)
if spikes_data.waveforms_filtered is None:
data = dict(
waveforms=None,
channels=channels,
clusters=None,
cluster_colors=None,
clusters_selected=clusters,
masks=None,
geometrical_positions=None,
autozoom=autozoom,
keep_order=wizard,
)
return data
_, nsamples, nchannels = spikes_data.waveforms_filtered.shape
# Find spikes to display and load the waveforms.
if len(clusters) > 0:
spikes_selected, waveforms = spikes_data.load_waveforms(
clusters=clusters,
count=USERPREF['waveforms_nspikes_max_expected'])
else:
spikes_selected = []
# Bake the waveform data.
if len(spikes_selected) > 0:
waveforms = convert_dtype(waveforms, np.float32)
if spikes_data.masks is not None:
masks = spikes_data.masks[spikes_selected,
0:fetdim * nchannels:fetdim]
else:
masks = None
else:
waveforms = np.zeros((0, nsamples, nchannels), dtype=np.float32)
masks = np.ones((0, nchannels), dtype=np.float32)
if masks is None:
masks = np.ones((len(spikes_selected), nchannels), dtype=np.float32)
spike_clusters = spike_clusters[spikes_selected]
channel_positions = np.array([
channels_data[ch].position
if channels_data[ch].position is not None else (0., ch)
for ch in channels
],
dtype=np.float32)
# Pandaize
waveforms = pandaize(waveforms, spikes_selected)
spike_clusters = pandaize(spike_clusters, spikes_selected)
masks = pandaize(masks, spikes_selected)
cluster_colors = pandaize(cluster_colors, clusters)
data = dict(
waveforms=waveforms,
channels=channels,
clusters=spike_clusters,
cluster_colors=cluster_colors,
clusters_selected=clusters,
masks=masks,
geometrical_positions=channel_positions,
autozoom=autozoom,
keep_order=wizard,
)
return data
def get_featureview_data(exp,
clusters=[],
channel_group=0,
clustering='main',
nspikes_bg=None,
autozoom=None,
alpha_selected=.75,
alpha_background=.25,
normalization=None,
time_unit='second'):
clusters = np.array(clusters)
# TODO: add spikes=None and spikes_bg=None
fetdim = exp.application_data.spikedetekt.n_features_per_channel
channels = exp.channel_groups[channel_group].channel_order
clusters_data = getattr(exp.channel_groups[channel_group].clusters,
clustering)
spikes_data = exp.channel_groups[channel_group].spikes
channels_data = exp.channel_groups[channel_group].channels
nchannels = len(channels_data)
spike_clusters = getattr(spikes_data.clusters, clustering)[:]
# cluster_colors = clusters_data.color[clusters]
# get colors from application data:
cluster_colors = pd.Series(
[_get_color(clusters_data, cl) for cl in clusters], index=clusters)
# cluster_colors = pd.Series([
# next_color(cl)
# if cl in clusters_data else 1
# for cl in clusters], index=clusters)
if len(clusters) > 0:
# TODO: put fraction in user parameters
spikes_selected, fm = spikes_data.load_features_masks(
clusters=clusters)
else:
spikes_selected = []
fm = np.zeros((0, spikes_data.features_masks.shape[1], 2),
dtype=spikes_data.features_masks.dtype)
fm = np.atleast_3d(fm)
features = fm[:, :, 0]
nextrafet = features.shape[1] - nchannels * fetdim
if fm.shape[2] > 1:
masks = fm[:, ::fetdim, 1]
else:
masks = None
nspikes = features.shape[0]
spiketimes_all = spikes_data.concatenated_time_samples[:]
spiketimes = spiketimes_all[spikes_selected]
spike_clusters = spike_clusters[spikes_selected]
freq = exp.application_data.spikedetekt.sample_rate
duration = spikes_data.concatenated_time_samples[
len(spikes_data.concatenated_time_samples) - 1] * 1. / freq
spikes_bg, features_bg = spikes_data.load_features_masks_bg()
features_bg = np.atleast_3d(features_bg)
features_bg = features_bg[:, :, 0].copy()
spiketimes_bg = spiketimes_all[spikes_bg]
# Add extra feature for time is necessary.
if nextrafet == 0:
features = np.hstack((features, np.ones((features.shape[0], 1))))
features_bg = np.hstack((features_bg, np.ones(
(features_bg.shape[0], 1))))
nextrafet = 1
# Normalize features.
def _find_max(x):
if x.size == 0:
return 1.
return np.max(np.abs(x))
c = (normalization or
(1. / _find_max(features_bg[:, :-nextrafet]))) if nspikes > 0 else 1.
features[:, :-nextrafet] *= c
features_bg[:, :-nextrafet] *= c
# Normalize extra features except time.
for i in range(features_bg.shape[1] - nextrafet - 1,
features_bg.shape[1] - 1):
c = (1. / _find_max(features_bg[:, i])) if nspikes > 0 else 1.
features[:, i] *= c
features_bg[:, i] *= c
# Normalize time.
if features.size > 0:
features[:, -1] = spiketimes
features[:, -1] *= (1. / (duration * freq))
features[:, -1] = 2 * features[:, -1] - 1
features_bg[:, -1] = spiketimes_bg
features_bg[:, -1] *= (1. / (duration * freq))
features_bg[:, -1] = 2 * features_bg[:, -1] - 1
# Pandaize
features = pandaize(features, spikes_selected)
features_bg = pandaize(features_bg, spikes_bg)
if masks is not None:
masks = pandaize(masks, spikes_selected)
spiketimes = pandaize(spiketimes, spikes_selected)
spike_clusters = pandaize(spike_clusters, spikes_selected)
cluster_colors = pandaize(cluster_colors, clusters)
# nextrafet = features.shape[1] - fetdim * nchannels
data = dict(
features=features,
features_background=features_bg,
masks=masks,
spiketimes=spiketimes,
clusters=spike_clusters,
clusters_selected=clusters,
cluster_colors=cluster_colors,
nchannels=nchannels,
channels=channels,
fetdim=fetdim,
nextrafet=nextrafet,
freq=freq,
autozoom=autozoom,
duration=duration,
alpha_selected=alpha_selected,
alpha_background=alpha_background,
time_unit=time_unit,
)
return data
