def ripple_spiking_statisitics(ripple_events, data_directory,position_data, ep, bin_size):
#For inspecting the spiking profile for different SWR events..
#Useful for inspecting the no of ripple events passing a specific statistical criterion
from wrappers import loadSpikeData
from functions import computeAngularTuningCurves, findHDCells
spikes, shanks = loadSpikeData(data_directory)
tuning_curves = (computeAngularTuningCurves(spikes, position_data, ep, 61))
hd_idx, stats = findHDCells(tuning_curves)
ripple_events = ripple_events.values
spikes_hd = {k:spikes[k] for k in hd_idx}
#print(spikes_hd)
d = 1000000/2
ripple_epochs = make_Epochs((ep['start'].values[0] + (ripple_events[:,1]*1000000) - d), (ep['start'].values[0] + (ripple_events[:,1]*1000000) + d))
#print(ripple_epochs.as_units('ms'))
val_array = []
edges = []
for i in range(len(ripple_epochs)):
val = 0
for k in hd_idx:
spk = spikes_hd[k].restrict(ripple_epochs.iloc[[i]]).as_units('ms').index.values
val_emp, edge = np.histogram(spk, np.arange(ripple_epochs.as_units('ms').iloc[[i]].start.values[0], ripple_epochs.as_units('ms').iloc[[i]].end.values[0], bin_size))
print(len(np.nonzero(val_emp)))
val = val + val_emp
val_array.append(val)
edges.append(edge)
print(val_array)
return val_array, edges
position = loadPosition(data_directory, events, episodes)
wake_ep = loadEpoch(data_directory, 'wake', episodes)
sleep_ep = loadEpoch(data_directory, 'sleep')
# We can look at the position of the animal in 2d with a figure
figure()
plot(position['x'], position['z'])
show()
# Now we are going to compute the tuning curve for all neurons during exploration
# The process of making a tuning curve has been covered in main3_tuningcurves.py
# So here we are gonna use the function computeAngularTuningCurves from functions.py
from functions import computeAngularTuningCurves
tuning_curves = computeAngularTuningCurves(spikes, position['ry'], wake_ep, 60)
# And let's plot all the tuning curves in a polar plot
from pylab import *
figure()
for i, n in enumerate(tuning_curves.columns):
subplot(10, 10, i + 1, projection='polar')
plot(tuning_curves[n])
show()
# It's a bit dirty. Let's smooth the tuning curves ...
from functions import smoothAngularTuningCurves
tuning_curves = smoothAngularTuningCurves(tuning_curves, 10, 2)
# and plot it again
figure()
def ripple_Decoding(ripple_event, data_dir, epoch, lfp, save_pdf = False):
#Purpose: Decoding the ripple event to the Head-Direction angle data
'''
NOTE ON function arguments:
ripple_event : pandas Dataframe with info about ripple events in the following form : [start time, peak time, end time, Average Power, Instantanous Frequency of ripple].
data_dir : path to the directory where all the session data is stored.
epoch : name of epoch in which decoding is needed to be done.
lfp : Downasampled neuroseries Time Series with single channel. It can be obtained using loadLFP() function to obtain a time series from a single channel.
'''
from matplotlib.backends.backend_pdf import PdfPages
import itertools
bin_size_un = 20 #ms
bin_size_ov = 100 #ms
percent_overlap = 0.75
d1,d2 = plot_sharing_calc(bin_size_un, bin_size_ov, (500 * 1000))
ripple_event = ripple_event.values
#Loading the correct angular position wrt time from Positions.h5 file.
positions = load_Positions(data_directory, epoch)
ep = correct_Epochs(positions)
#Loading the spike and shank data
from wrappers import loadSpikeData, loadEpoch, loadPosition, loadXML
spikes, shank = loadSpikeData(data_dir)
n_channels, fs, shank_to_channel = loadXML(data_dir)
#Loading the wakefulness period data for decoding as it is required to calculate tuning curves and occupancy
pos_wake = load_Positions(data_directory, 'wake')
wake_ep = correct_Epochs(pos_wake)
#Computing Angular Tuning curves for the detection of head-direction cells using Reyleigh's test.
#NOTE: Use wakefulness position data to calculate tuning curves and occupancy
from functions import computeAngularTuningCurves
tuning_curves = computeAngularTuningCurves(spikes, pos_wake['angle'], wake_ep, 61) #spikes: spike data, angle : angle data from all the, no of bins for angles
#Defining correct HD Cells using Reileigh's test on angular tuning curves.
from functions import findHDCells
hd_idx, stat = findHDCells(tuning_curves)
tuning_curves = tuning_curves[hd_idx]
#To calculate prior using Poisson spiking assumption
from functions import decodeHD, decodeHD_overlap2
spikes_hd = {k:spikes[k] for k in hd_idx}
occupancy = np.histogram(pos_wake['angle'], np.linspace(0, 2*np.pi, 61), weights = np.ones_like(pos_wake['angle'])/float(len(pos_wake['angle'])))[0]
#Here, we will try to decode the HD angle during wake state to see whether the decoding priors like Ocuupancy map and tuning curves are correct or not.
wake_dec, prob_wake_dec = decodeHD(tuning_curves, spikes_hd, wake_ep, occupancy, 200)
posterior_wake = nts.Tsd(t = prob_wake_dec.index.values, d = prob_wake_dec.max(1).values, time_units = 'ms')
'''
figure()
subplot(2,1,1)
subplot(2,1,1).set_title('Bayesian Decoding during Wakefulness')
plot(pos_wake['angle'].restrict(wake_ep), label = 'True HD')
plot(wake_dec, label = 'Decoded HD(Bayesian)')
legend()
subplot(2,1,2)
subplot(2,1,2).set_title('Posterior Probability(Bayesian Decoding)')
matshow(prob_wake_dec)
legend()
show()
'''
print(ripple_event.shape)
#So far, we have calculated the general information that we need for decoding such as occupancy map and spikes of HD cells. They are
#important irrespective of the event we are trying to decode.
#Now, we will calculate the decoding and probable_angle(wrt time) for each ripple event
d = 500 * 1000
ripple_epochs1 = make_Epochs((ep['start'].values[0] + (ripple_event[:,1]*1000000) - d1), (ep['start'].values[0] + (ripple_event[:,1]*1000000) + d1)) #ripple_event is in seconds here...
ripple_epochs2 = make_Epochs((ep['start'].values[0] + (ripple_event[:,1]*1000000) - d2), (ep['start'].values[0] + (ripple_event[:,1]*1000000) + d2))
ripple_epochs = make_Epochs((ep['start'].values[0] + (ripple_event[:,1]*1000000) - d), (ep['start'].values[0] + (ripple_event[:,1]*1000000) + d))
proba_angle = [] #Since there are multiple events to decode, the proba_angle will now be a list of decodings for each of the ripple event
decoded = []
#e = 67
#signal = signal.as_units('s')
if save_pdf == True:
print("Initiating Ripple Plotting..")
with PdfPages("Ripple_plots_all.pdf") as Pdf:
#Some preprocessing to display LFP together...
low_cut = 200
high_cut = 400
frequency = 1250
signal = bandPass_filt(lfp_unfiltered = lfp, low_cut = low_cut, high_cut = high_cut, sampling_freq = frequency)
ts = np.arange(len(signal))/frequency
signal = nts.TsdFrame(ts, signal, time_units = 's')
for e in range(5):
#for e in range(len(ripple_epochs.index.values)):
#Decoding using Non-overlapping equal width bins
decoding, p_angle = decodeHD(tuning_curves, spikes_hd, ripple_epochs1[e], occupancy, bin_size_un)
#posterior_prob = nts.Tsd(t = p_angle.index.values, d = p_angle.max(1).values, time_units = 'ms')
p_mat = p_angle.values
entropy1 = -(p_angle*np.log2(p_angle)).sum(1)
#print(entropy1)
#entropy1 = nts.Tsd(t = p_angle.index.values, d = ent1, time_units = 'ms')
#Decoding using Overlapping bins
decoding2, p_angle2 = decodeHD_overlap2(tuning_curves, spikes_hd, ripple_epochs2[e], occupancy, bin_size_ov, percent_overlap = percent_overlap)
if type(decoding2).__name__ == 'NoneType':
continue
p_mat2 = p_angle2.values
entropy2 = -(p_angle2*np.log2(p_angle2)).sum(1)
#print(entropy2)
#posterior_prob2 = nts.Tsd(t = p_angle2.index.values, d = p_angle2.max(1).values, time_units = 'ms')
#entropy2 = nts.Tsd(t = p_angle2.index.values, d = [-p*math.log(p) for p in posterior_prob2.values], time_units = 'ms')
frac = (1/len(spikes_hd))
ind_dix = pd.DataFrame(hd_idx, index = [tuning_curves[k].idxmax() for k in hd_idx], columns = ["cell_id"])
#ind_dix = {tuning_curves[k].idxmax():k for k in hd_idx}
w = ind_dix.sort_index()
#print(hd_idx)
lfp_view = signal.restrict(ripple_epochs[e])
#lfp_view = signal[((ep.as_units('s')['start'].values[0]) + (ripple_event[int(e),0]) - resolution) : ((ep.as_units('s')['start'].values[0]) + (ripple_event[int(e),2]) + resolution)]
print("Plotting Figure "+ str(e+1)+".")
figure()
ax1 = plt.subplot(4,1,1)
ax1.set_title('LFP Bandpassed('+ str(low_cut) + '-' + str(high_cut) + 'Hz)')
ax1.set_ylabel('mV')
ax1.plot(lfp_view)
ax1.plot(lfp.restrict(ripple_epochs[e]))
ax2 = plt.subplot(4,1,2, sharex = ax1)
[[ax2.axvline(_x, linewidth=1, color='r', ymin = i*frac, ymax = (i+1)*frac) for _x in spikes_hd[col].restrict(ripple_epochs[e]).index.values] for col,i in itertools.zip_longest(w['cell_id'].values, np.arange(len(spikes_hd)))]
#[ax2.axhline((2*i +1)*frac/2, linewidth=1, color='black') for i in range(len(spikes_hd)+1)]
ax2.axis('tight')
ax2.set_yticks([(2*i + 1)*frac/2 for i in range(len(hd_idx)+1)])
ax2.set_yticklabels([str('%.2f'%i) for i in w.index.values])
#ax2.set_title('Rastor Plot')
ax2.set_ylabel('HD Neuron acc. Preferred Angle')
ax3 = plt.subplot(4,1,3, sharex = ax2)
#ax3.set_title('Bayesian Decoding')
ax3.plot(positions['angle'].restrict(ripple_epochs[e]), label = 'True HD')
ax3.plot(decoding, label = 'Decoded HD')
ax3.set_ylabel('Decoded Angle(Radians)')
ax3.plot(decoding2, label = 'Decoded HD(overlapping)')
#ax3.axvline(x = (ripple_epochs.iloc[[e]])['start'].values[0], color = 'green', label = 'Starting point of SWR')
#ax3.axvline(x = (ripple_epochs.iloc[[e]])['end'].values[0], color = 'red', label = 'End point of SWR')
ax3.legend()
ax4 = plt.subplot(4,1,4)
#ax4.set_title('Posterior Prob(Bayesian)')
ax4.plot(entropy1, label = 'Entropy')
ax4.plot(entropy2, label = 'Entropy(Overlapping)')
ax4.set_ylabel('Entropy of Decoding')
#ax4.axvline(x = (ripple_epochs.iloc[[e]])['start'].values[0], color = 'green', label = 'Starting point of SWR')
#ax4.axvline(x = (ripple_epochs.iloc[[e]])['end'].values[0], color = 'red', label = 'End point of SWR')
ax4.legend()
mng = plt.get_current_fig_manager()
mng.window.showMaximized()
plt.pause(0.2)
try:
Pdf.savefig(dpi = 600)
except ValueError:
continue
#legend()
#show()
plt.close()
print("All Plotted. Done!")
#sys.exit()
entropy_un = []
entropy_b = []
decoded2 = []
decoded = []
print("trying decoding")
for i in range(len(ripple_epochs2)):
#decoding, p_angle = decodeHD(tuning_curves, spikes_hd, ripple_epochs1[i], occupancy, bin_size_un)
#decoded.append(decoding)
#decoding, p_angle = decodeHD(tuning_curves, spikes_hd, ripple_epochs.iloc[[i]], occupancy, 20)
decoding2, p_angle2 = decodeHD_overlap2(tuning_curves, spikes_hd, ripple_epochs2[i], occupancy, bin_size_ov, percent_overlap = percent_overlap)
print(len(decoding2))
decoded2.append(decoding2)
#entropy1 = -(p_angle*np.log2(p_angle)).sum(1)
#entropy_un.append(entropy1)
#decoded.append(decoding)
#proba_angle.append(p_angle)
#entropy1 = -(p_angle*np.log2(p_angle)).sum(1)
#entropy_un.append(entropy1)
p_mat2 = p_angle2.values
entropy2 = -(p_angle2*np.log2(p_angle2)).sum(1)
entropy_b.append(entropy2)
print("Done with "+ str(i))
return entropy_b, decoded2