# Get all the units from the hdf5 file
units = hf5.list_nodes('/sorted_units')
# Delete and remake a directory for storing the plots of the unit waveforms (if it exists)
try:
shutil.rmtree("unit_waveforms_plots", ignore_errors=True)
except:
pass
os.mkdir("unit_waveforms_plots")
# Now plot the waveforms from the units in this directory one by one
for unit in range(len(units)):
waveforms = units[unit].waveforms[:]
x = np.arange(waveforms.shape[1] / 10) + 1
fig, ax = blech_waveforms_datashader.waveforms_datashader(waveforms, x)
ax.set_xlabel('Sample (30 samples per ms)')
ax.set_ylabel('Voltage (microvolts)')
ax.set_title('Unit %i, total waveforms = %i' % (unit, waveforms.shape[0]) +
'\n' + 'Electrode: %i, Single Unit: %i, RSU: %i, FS: %i' %
(hf5.root.unit_descriptor[unit]['electrode_number'],
hf5.root.unit_descriptor[unit]['single_unit'],
hf5.root.unit_descriptor[unit]['regular_spiking'],
hf5.root.unit_descriptor[unit]['fast_spiking']))
fig.savefig('./unit_waveforms_plots/Unit%i.png' % (unit))
plt.close("all")
# Also plot the mean and SD for every unit - downsample the waveforms 10 times to remove effects of upsampling during de-jittering
fig = plt.figure()
plt.plot(x, np.mean(waveforms[:, ::10], axis=0), linewidth=4.0)
plt.fill_between(x,
# Get all the units from the hdf5 file
units = hf5.list_nodes('/sorted_units')
# Delete and remake a directory for storing the plots of the unit waveforms (if it exists)
try:
shutil.rmtree("unit_waveforms_plots", ignore_errors = True)
except:
pass
os.mkdir("unit_waveforms_plots")
# Now plot the waveforms from the units in this directory one by one
for unit in range(len(units)):
waveforms = units[unit].waveforms[:]
x = np.arange(waveforms.shape[1]/10) + 1
fig, ax = blech_waveforms_datashader.waveforms_datashader(waveforms, x)
ax.set_xlabel('Sample (30 samples per ms)')
ax.set_ylabel('Voltage (microvolts)')
ax.set_title('Unit %i, total waveforms = %i' % (unit, waveforms.shape[0]) + '\n' + 'Electrode: %i, Single Unit: %i, RSU: %i, FS: %i' % (hf5.root.unit_descriptor[unit]['electrode_number'], hf5.root.unit_descriptor[unit]['single_unit'], hf5.root.unit_descriptor[unit]['regular_spiking'], hf5.root.unit_descriptor[unit]['fast_spiking']))
fig.savefig('./unit_waveforms_plots/Unit%i.png' % (unit))
plt.close("all")
# Also plot the mean and SD for every unit - downsample the waveforms 10 times to remove effects of upsampling during de-jittering
fig = plt.figure()
plt.plot(x, np.mean(waveforms[:, ::10], axis = 0), linewidth = 4.0)
plt.fill_between(x, np.mean(waveforms[:, ::10], axis = 0) - np.std(waveforms[:, ::10], axis = 0), np.mean(waveforms[:, ::10], axis = 0) + np.std(waveforms[:, ::10], axis = 0), alpha = 0.4)
plt.xlabel('Sample (30 samples per ms)')
plt.ylabel('Voltage (microvolts)')
plt.title('Unit %i, total waveforms = %i' % (unit, waveforms.shape[0]) + '\n' + 'Electrode: %i, Single Unit: %i, RSU: %i, FS: %i' % (hf5.root.unit_descriptor[unit]['electrode_number'], hf5.root.unit_descriptor[unit]['single_unit'], hf5.root.unit_descriptor[unit]['regular_spiking'], hf5.root.unit_descriptor[unit]['fast_spiking']))
fig.savefig('./unit_waveforms_plots/Unit%i_mean_sd.png' % (unit))
plt.close("all")
g.fit(data)
# Show the cluster plots if the solution converged
if g.converged_:
split_predictions = g.predict(data)
x = np.arange(len(spike_waveforms[0])/10) + 1
for cluster in range(n_clusters):
split_points = np.where(split_predictions == cluster)[0]
# plt.figure(cluster)
slices_dejittered = spike_waveforms[this_cluster, :] # Waveforms and times from the chosen cluster
times_dejittered = spike_times[this_cluster]
times_dejittered = times_dejittered[split_points] # Waveforms and times from the chosen split of the chosen cluster
ISIs = np.ediff1d(np.sort(times_dejittered))/30.0
violations1 = 100.0*float(np.sum(ISIs < 1.0)/split_points.shape[0])
violations2 = 100.0*float(np.sum(ISIs < 2.0)/split_points.shape[0])
fig, ax = blech_waveforms_datashader.waveforms_datashader(slices_dejittered[split_points, :], x)
# plt.plot(x-15, slices_dejittered[split_points, :].T, linewidth = 0.01, color = 'red')
ax.set_xlabel('Sample (30 samples per ms)')
ax.set_ylabel('Voltage (microvolts)')
ax.set_title("Split Cluster{:d}, 2ms ISI violations={:.1f} percent".format(cluster, violations2) + "\n" + "1ms ISI violations={:.1f}%, Number of waveforms={:d}".format(violations1, split_points.shape[0]))
else:
print("Solution did not converge - try again with higher number of iterations or lower convergence criterion")
continue
plt.show()
# Ask the user for the split clusters they want to choose
chosen_split = easygui.multchoicebox(msg = 'Which split cluster do you want to choose? Hit cancel to exit', choices = tuple([str(i) for i in range(n_clusters)]))
try:
chosen_split = int(chosen_split[0])
except:
continue
split_predictions = g.predict(data)
x = np.arange(len(spike_waveforms[0]) / 10) + 1
for cluster in range(n_clusters):
split_points = np.where(split_predictions == cluster)[0]
# plt.figure(cluster)
slices_dejittered = spike_waveforms[
this_cluster, :] # Waveforms and times from the chosen cluster
times_dejittered = spike_times[this_cluster]
times_dejittered = times_dejittered[
split_points] # Waveforms and times from the chosen split of the chosen cluster
ISIs = np.ediff1d(np.sort(times_dejittered)) / 30.0
violations1 = 100.0 * float(
np.sum(ISIs < 1.0) / split_points.shape[0])
violations2 = 100.0 * float(
np.sum(ISIs < 2.0) / split_points.shape[0])
fig, ax = blech_waveforms_datashader.waveforms_datashader(
slices_dejittered[split_points, :], x)
# plt.plot(x-15, slices_dejittered[split_points, :].T, linewidth = 0.01, color = 'red')
ax.set_xlabel('Sample (30 samples per ms)')
ax.set_ylabel('Voltage (microvolts)')
ax.set_title(
"Split Cluster{:d}, 2ms ISI violations={:.1f} percent".
format(cluster, violations2) + "\n" +
"1ms ISI violations={:.1f}%, Number of waveforms={:d}".
format(violations1, split_points.shape[0]))
else:
print(
"Solution did not converge - try again with higher number of iterations or lower convergence criterion"
)
continue
plt.show()
# Show the cluster plots if the solution converged
if g.converged_:
split_predictions = g.predict(data)
x = np.arange(len(spike_waveforms[0])) + 1
#fig, ax = gen_square_subplots(n_clusters,sharex=True,sharey=True)
for cluster in range(n_clusters):
split_points = np.where(split_predictions == cluster)[0]
# Waveforms and times from the chosen cluster
slices_dejittered = spike_waveforms[this_cluster, :]
times_dejittered = spike_times[this_cluster]
# Waveforms and times from the chosen split of the chosen cluster
times_dejittered = times_dejittered[split_points]
ISIs = np.ediff1d(np.sort(times_dejittered))/30.0
violations1 = 100.0*float(np.sum(ISIs < 1.0)/split_points.shape[0])
violations2 = 100.0*float(np.sum(ISIs < 2.0)/split_points.shape[0])
fig, ax = blech_waveforms_datashader.waveforms_datashader(\
slices_dejittered[split_points, :], x, downsample = False)
ax.set_xlabel('Sample (30 samples per ms)')
ax.set_ylabel('Voltage (microvolts)')
print_str = (f'\nSplit Cluster {cluster} \n'
f'{violations2:.1f} % (<2ms),'
f'{violations1:.1f} % (<1ms),'
f'{split_points.shape[0]} total waveforms. \n')
ax.set_title(print_str)
else:
print("Solution did not converge "\
"- try again with higher number of iterations "\
"or lower convergence criterion")
continue
plt.show()
f'clusters{i+2}/predictions.npy',
predictions)
# Create file, and plot spike waveforms for the different clusters.
# Plot 10 times downsampled dejittered/smoothed waveforms.
# Additionally plot the ISI distribution of each cluster
os.mkdir(f'./Plots/{electrode_num:02}/{i+2}_clusters_waveforms_ISIs')
x = np.arange(len(slices_dejittered[0])) + 1
for cluster in range(i + 2):
cluster_points = np.where(predictions[:] == cluster)[0]
if len(cluster_points) > 0:
fig, ax = \
blech_waveforms_datashader.waveforms_datashader(\
slices_dejittered[cluster_points, :],
x,
downsample = False,
dir_name = "datashader_temp_el" + str(electrode_num))
ax.set_xlabel('Sample ({:d} samples per ms)'.\
format(int(sampling_rate/1000)))
ax.set_ylabel('Voltage (microvolts)')
ax.set_title('Cluster%i' % cluster)
fig.savefig(f'./Plots/{electrode_num:02}/'\
f'{i+2}_clusters_waveforms_ISIs/Cluster{cluster}_waveforms')
plt.close("all")
fig = plt.figure()
cluster_times = times_dejittered[cluster_points]
ISIs = np.ediff1d(np.sort(cluster_times))
ISIs = ISIs / 30.0
max_ISI_val = 20
# Additionally plot the ISI distribution of each cluster
os.mkdir('./Plots/%i/Plots/%i_clusters_waveforms_ISIs' %
(electrode_num, i + 2))
x = np.arange(len(slices_dejittered[0]) / 10) + 1
for cluster in range(i + 2):
cluster_points = np.where(predictions[:] == cluster)[0]
#for point in cluster_points:
# plot_wf = np.zeros(len(slices_dejittered[0])/10)
# for time in range(len(slices_dejittered[point])/10):
# plot_wf[time] = slices_dejittered[point, time*10]
# plt.plot(x-15, plot_wf, linewidth = 0.1, color = 'red')
# plt.hold(True)
# plt.plot(x - int((sampling_rate/1000.0)*spike_snapshot_before), slices_dejittered[cluster_points, ::10].T, linewidth = 0.01, color = 'red')
fig, ax = blech_waveforms_datashader.waveforms_datashader(
slices_dejittered[cluster_points, :],
x,
dir_name="datashader_temp_el" + str(electrode_num))
ax.set_xlabel('Sample ({:d} samples per ms)'.format(
int(sampling_rate / 1000)))
ax.set_ylabel('Voltage (microvolts)')
ax.set_title('Cluster%i' % cluster)
fig.savefig(
'./Plots/%i/Plots/%i_clusters_waveforms_ISIs/Cluster%i_waveforms' %
(electrode_num, i + 2, cluster))
plt.close("all")
fig = plt.figure()
cluster_times = times_dejittered[cluster_points]
ISIs = np.ediff1d(np.sort(cluster_times))
ISIs = ISIs / 30.0
plt.hist(ISIs,
# Create file, and plot spike waveforms for the different clusters. Plot 10 times downsampled dejittered/smoothed waveforms.
# Additionally plot the ISI distribution of each cluster
os.mkdir('./Plots/%i/Plots/%i_clusters_waveforms_ISIs' % (electrode_num, i+2))
x = np.arange(len(slices_dejittered[0])/10) + 1
for cluster in range(i+2):
cluster_points = np.where(predictions[:] == cluster)[0]
#for point in cluster_points:
# plot_wf = np.zeros(len(slices_dejittered[0])/10)
# for time in range(len(slices_dejittered[point])/10):
# plot_wf[time] = slices_dejittered[point, time*10]
# plt.plot(x-15, plot_wf, linewidth = 0.1, color = 'red')
# plt.hold(True)
# plt.plot(x - int((sampling_rate/1000.0)*spike_snapshot_before), slices_dejittered[cluster_points, ::10].T, linewidth = 0.01, color = 'red')
fig, ax = blech_waveforms_datashader.waveforms_datashader(slices_dejittered[cluster_points, :], x, dir_name = "datashader_temp_el" + str(electrode_num))
ax.set_xlabel('Sample ({:d} samples per ms)'.format(int(sampling_rate/1000)))
ax.set_ylabel('Voltage (microvolts)')
ax.set_title('Cluster%i' % cluster)
fig.savefig('./Plots/%i/Plots/%i_clusters_waveforms_ISIs/Cluster%i_waveforms' % (electrode_num, i+2, cluster))
plt.close("all")
fig = plt.figure()
cluster_times = times_dejittered[cluster_points]
ISIs = np.ediff1d(np.sort(cluster_times))
ISIs = ISIs/30.0
plt.hist(ISIs, bins = [0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, np.max(ISIs)])
plt.xlim([0.0, 10.0])
plt.title("2ms ISI violations = %.1f percent (%i/%i)" %((float(len(np.where(ISIs < 2.0)[0]))/float(len(cluster_times)))*100.0, len(np.where(ISIs < 2.0)[0]), len(cluster_times)) + '\n' + "1ms ISI violations = %.1f percent (%i/%i)" %((float(len(np.where(ISIs < 1.0)[0]))/float(len(cluster_times)))*100.0, len(np.where(ISIs < 1.0)[0]), len(cluster_times)))
fig.savefig('./Plots/%i/Plots/%i_clusters_waveforms_ISIs/Cluster%i_ISIs' % (electrode_num, i+2, cluster))
plt.close("all")