def test_create_ts_time_units(self):
"""
internally data are stored as us
"""
a = np.random.randint(0, 1000, 100)
a.sort()
ts = nts.Ts(a / 1000, time_units='ms')
np.testing.assert_array_almost_equal_nulp(ts.index.values, a)
ts = nts.Ts(a / 1000000, time_units='s')
# np.testing.assert_array_almost_equal_nulp(ts.index.values, a.astype(np.int64), nulp=100)
self.assertTrue(np.all(ts.index.values >= a.astype(np.int64) - 1))
self.assertTrue(np.all(ts.index.values <= a.astype(np.int64) + 1))
def loadSpikeData(path, index):
# units shoud be the value to convert in s
import scipy.io
import neuroseries as nts
spikedata = scipy.io.loadmat(path)
shank = spikedata['shank']
shankIndex = np.where(shank == index)[0]
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(spikedata['S'][0][0][0][i][0][0][0][1][0][0][2], time_units = 's')
a = spikes[list(spikes.keys())[0]].as_units('s').index.values
if ((a[-1]-a[0])/60.)/60. > 20. : # VERY BAD
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(spikedata['S'][0][0][0][i][0][0][0][1][0][0][2]*0.0001, time_units = 's')
return spikes, shank
def test_realign_right(self, data_class):
d_a = self.mat_data_right['d_a']
d_a = d_a.reshape((len(d_a), ))
t_a = data_class(self.mat_data_right['t_a'].astype(np.int64), d_a)
t_b = nts.Ts(self.mat_data_right['t_b'].astype(np.int64))
t_closest = t_a.realign(t_b)
dt = self.mat_data_right['d_closest'].ravel()
self.assertTrue((t_closest.values.ravel() != dt).sum() < 10)
np.testing.assert_array_almost_equal_nulp(t_closest.values.ravel(), dt)
def test_create_ts_time_units_double():
"""
conversion of time units from floating point type
"""
a = np.floor(np.random.rand(100) * 1000000)
a.sort()
ts = nts.Ts(a.copy(), time_units='ms')
# noinspection PyTypeChecker
np.testing.assert_array_almost_equal_nulp(ts.index.values / 1000, a)
def test_create_ts_from_double(self):
"""
data get converted to ts and back fine
"""
a = np.floor(np.random.rand(100)*1000000)
a.sort()
ts = nts.Ts(a)
self.assertIs(ts.index.dtype, np.dtype(np.int64), msg='index type is not int64')
np.testing.assert_array_almost_equal_nulp(a, ts.index.values)
def test_realign_wrong_units(self, data_class):
d_a = self.mat_data1['d_a']
d_a = d_a.reshape((len(d_a), ))
t_a = data_class(self.mat_data1['t_a'].astype(np.int64), d_a)
t_b = nts.Ts(self.mat_data1['t_b'].astype(np.int64))
# noinspection PyUnusedLocal
t_closest = 1
with self.assertRaises(ValueError):
t_closest = t_a.realign(t_b, align='banana')
self.assertTrue(t_closest)
def test_create_ts_from_non_sorted():
"""
if ts are not sorted, a warning should be returned and the timestamps sorted for you
"""
a = np.random.randint(0, 1000, 100)
# with self.assertWarns(UserWarning):
# ts = nts.Ts(a)
ts = nts.Ts(a)
np.testing.assert_array_almost_equal_nulp(np.sort(a), ts.index.values)
def loadRipples(path):
# 0 : debut
# 1 : milieu
# 2 : fin
# 3 : amplitude nombre de sd au dessus de bruit
# 4 : frequence instantan
import neuroseries as nts
ripples = np.genfromtxt(path + '/' + path.split("/")[-1] + '.sts.RIPPLES')
return (nts.IntervalSet(ripples[:, 0], ripples[:, 2], time_units='s'),
nts.Ts(ripples[:, 1], time_units='s'))
def test_times_units_ts():
"""
tests the units calling of times
"""
a = np.random.randint(0, 10000000, 100)
a.sort()
ts = nts.Ts(a)
np.testing.assert_array_almost_equal_nulp(a, ts.times('us'))
np.testing.assert_array_almost_equal_nulp(a / 1000., ts.times('ms'))
np.testing.assert_array_almost_equal_nulp(a / 1.e6, ts.times('s'))
def test_create_ts_wrong_units(self):
"""
if the units are unsupported it should raise ValueError
"""
a = np.random.randint(0, 10000000, 100)
a.sort()
# noinspection PyUnusedLocal
ts = 1
with self.assertRaises(ValueError):
ts = nts.Ts(a, time_units='min')
self.assertTrue(ts)
def ripples():
ripples_ = scipy.io.loadmat(
f"{bk.load.session}-RippleFiring.mat")["ripples"]["allsws"][0][0]
# ripples_ = pd.DataFrame(data = ripples,columns=['start','peak','stop'])
columns = ["start", "peak", "stop"]
ripples = {}
for i, c in zip(range(ripples_.shape[1]), columns):
ripples.update({c: nts.Ts(ripples_[:, i], time_units="s")})
return ripples
def ripples():
ripples_ = scipy.io.loadmat(
f'{bk.load.session}-RippleFiring.mat')['ripples']['allsws'][0][0]
# ripples_ = pd.DataFrame(data = ripples,columns=['start','peak','stop'])
columns = ['start', 'peak', 'stop']
ripples = {}
for i, c in zip(range(ripples_.shape[1]), columns):
ripples.update({c: nts.Ts(ripples_[:, i], time_units='s')})
return ripples
def transitions_times(states,epsilon = 1,verbose = False):
'''
states : dict of nts.Interval_Set
This function compute transition in between Intervals in a dict.
It returns a new dict with intervals and when the transition occurs
epsilon : tolerance time delay between state
This function does NOT WORK for triple transitions (ex : sws/rem/sws) ...
'''
import itertools
empty_state = []
for state in states:
if len(states[state]) == 0:
empty_state.append(state)
continue
states[state] = states[state].drop_short_intervals(1)
for i in empty_state: del states[i]
transitions_intervals = {}
transitions_timing = {}
for items in itertools.permutations(states.keys(),2):
# states[items[0]] = states[items[0]].drop_short_intervals(1)
# states[items[1]] = states[items[1]].drop_short_intervals(1)
if verbose: print('Looking at transition from',items[0],' to ',items[1])
end = nts.Ts(np.array(states[items[0]].end + (epsilon * 1_000_000)+1))
in_next_epoch = states[items[1]].in_interval(end)
transitions_intervals.update({items:[]})
transitions_timing.update({items:[]})
for n,t in enumerate(in_next_epoch):
if np.isnan(t): continue
start = states[items[0]].iloc[n].start
trans = int(np.mean([states[items[0]].iloc[n].end,states[items[1]].iloc[int(t)].start]))
end = states[items[1]].iloc[int(t)].end
transitions_intervals[items].append([start,end])
transitions_timing[items].append(trans)
if not transitions_timing[items] == []:
transitions_intervals[items] = np.array(transitions_intervals[items])
transitions_intervals[items] = nts.IntervalSet(transitions_intervals[items][:,0],transitions_intervals[items][:,1],force_no_fix = True)
transitions_timing[items] = nts.Ts(t = np.array(transitions_timing[items]))
return transitions_intervals,transitions_timing
def test_create_ts(self):
"""
calling convention
ts = nts.Ts(a)
ts is an instance of pd.DataFrame
"""
a = np.random.randint(0, 10000000, 100)
a.sort()
ts = nts.Ts(a)
self.assertIsInstance(ts, pd.Series, msg="ts doesn't return DataFrame")
self.assertIsInstance(ts.index.values, np.ndarray,
msg="ts doesn't return array values")
self.assertIs(ts.index.dtype, np.dtype(np.int64), msg='index type is not int64')
np.testing.assert_array_almost_equal_nulp(a, ts.index.values)
def loadUFOs(path):
"""
Name of the file should end with .evt.py.ufo
"""
import os
name = path.split("/")[-1]
files = os.listdir(path)
filename = os.path.join(path, name + '.evt.py.ufo')
if name + '.evt.py.ufo' in files:
tmp = np.genfromtxt(path + '/' + name + '.evt.py.ufo')[:, 0]
ripples = tmp.reshape(len(tmp) // 3, 3) / 1000
else:
print("No ufo in ", path)
sys.exit()
return (nts.IntervalSet(ripples[:, 0], ripples[:, 2], time_units='s'),
nts.Ts(ripples[:, 1], time_units='s'))
def freezing_video(video_path, output_file, tf, freezing_intervals):
"""
video_path : path to the video to be displaying
outputfile : path to the video to written
tf : vector of time containing timing of each frame
freezing intervals : Intervals when the animal is freezing (as nts.Interval_Set)
"""
import cv2
if os.path.exists(output_file):
print(output_file, 'already exist, please delete manually')
return
print(video_path)
tf = nts.Ts(tf, time_units='s')
freezing_frames = np.where(freezing_intervals.in_interval(tf) >= 0)[0]
fs = 1 / scipy.stats.mode(np.diff(tf.as_units('s').index)).mode[0]
cap = cv2.VideoCapture(video_path)
frame_width = int(cap.get(3))
frame_height = int(cap.get(4))
nf = 0
out = cv2.VideoWriter(output_file,
cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'), fs,
(frame_width, frame_height))
while True:
ret, frame = cap.read()
if ret == True:
if nf in freezing_frames:
frame = cv2.circle(frame, (25, 25), 10, (0, 0, 255), 20)
cv2.imshow(video_path, frame)
out.write(frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
nf += 1
else:
break
cap.release()
out.release()
return True
def test_realign(self, data_class):
"""
first simple realign case
"""
d_a = self.mat_data1['d_a']
d_a = d_a.reshape((len(d_a),))
t_a = data_class(self.mat_data1['t_a'].astype(np.int64), d_a)
t_b = nts.Ts(self.mat_data1['t_b'].astype(np.int64))
t_closest = t_a.realign(t_b, align='closest')
dt = self.mat_data1['d_closest'].reshape((len(self.mat_data1['d_closest'],)))
self.assertTrue((t_closest.values.ravel() != dt).sum() < 10)
np.testing.assert_array_almost_equal_nulp(t_closest.values.ravel(), dt)
t_next = t_a.realign(t_b, align='next')
dt = self.mat_data1['d_next'].reshape((len(self.mat_data1['d_next'],)))
np.testing.assert_array_almost_equal_nulp(t_next.values.ravel(), dt)
t_prev = t_a.realign(t_b, align='prev')
dt = self.mat_data1['d_prev'].reshape((len(self.mat_data1['d_prev'],)))
np.testing.assert_array_almost_equal_nulp(t_prev.values.ravel(), dt)
def complextrain(epochs, num_ofneurons):
"""
This function simulate neural activity during several time epochs of
activity or innactivity
Parameters
----------
epochs : a neuroseries.interval_set.IntervalSet
an IntervalSet with the start and end of a time period and a label
indicating if the neuron was "active" or "inactive"
num_ofneurons : int
desired number of neurons to simulate
Returns
-------
complex_spikes : dict
a dictionary with simulated spike times as values and indices as keys
"""
complex_spikes = {neuron:emptyList for neuron, emptyList in \
zip ([*range(num_ofneurons)], [[] for i in range(num_ofneurons)])}
for i in epochs.index:
if epochs.loc[i].label == 'active':
spikes = sim.traingeneration(num_ofneurons,
epochs.loc[i].start,
epochs.loc[i].end,
factor=1)
elif epochs.loc[i].label == 'inactive':
spikes = sim.traingeneration(num_ofneurons,
epochs.loc[i].start,
epochs.loc[i].end,
factor=0.25)
else:
print("wrong label for epochs")
for n in complex_spikes.keys():
complex_spikes[n].extend(spikes[n])
for n in complex_spikes.keys():
array = np.sort(np.asarray(complex_spikes[n]))
complex_spikes[n] = nts.Ts(array, time_units='us')
return complex_spikes
(np.diff(obins) / 2).flatten(),
columns=neurons)
for k in neurons:
spks = spikes[k].as_units('ms').index.values
spike_counts[k] = histo(spks, tbins)
rates_swr.append(np.sqrt(spike_counts / (bin_size)))
####################################################################################################################
# RANDOM
####################################################################################################################
# BINNING
rnd_tsd = nts.Ts(t=np.sort(
np.hstack([
np.random.randint(sws_ep.loc[j, 'start'] +
500000, sws_ep.loc[j, 'end'] + 500000,
np.maximum(1, n_ex // len(sws_ep)))
for j in sws_ep.index
])))
if len(rnd_tsd) > n_ex:
rnd_tsd = rnd_tsd[0:n_ex]
rates_rnd = []
tmp3 = rnd_tsd.index.values / 1000
for j, t in enumerate(tmp3):
tbins = t + obins
spike_counts = pd.DataFrame(index=obins[:, 0] +
(np.diff(obins) / 2).flatten(),
columns=neurons)
for k in neurons:
spks = spikes[k].as_units('ms').index.values
spike_counts[k] = histo(spks, tbins)
def loadSpikeData(path, index=None, fs=20000):
"""
if the path contains a folder named /Analysis,
the script will look into it to load either
- SpikeData.mat saved from matlab
- SpikeData.h5 saved from this same script
if not, the res and clu file will be loaded
and an /Analysis folder will be created to save the data
Thus, the next loading of spike times will be faster
Notes :
If the frequency is not givne, it's assumed 20kH
Args:
path : string
Returns:
dict, array
"""
if not os.path.exists(path):
print("The path " + path + " doesn't exist; Exiting ...")
sys.exit()
new_path = os.path.join(path, 'Analysis/')
if os.path.exists(new_path):
new_path = os.path.join(path, 'Analysis/')
files = os.listdir(new_path)
if 'SpikeData.mat' in files:
spikedata = scipy.io.loadmat(new_path + 'SpikeData.mat')
shank = spikedata['shank'] - 1
if index is None:
shankIndex = np.arange(len(shank))
else:
shankIndex = np.where(shank == index)[0]
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(
spikedata['S'][0][0][0][i][0][0][0][1][0][0][2],
time_units='s')
a = spikes[0].as_units('s').index.values
if ((a[-1] - a[0]) / 60.) / 60. > 20.: # VERY BAD
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(
spikedata['S'][0][0][0][i][0][0][0][1][0][0][2] *
0.0001,
time_units='s')
return spikes, shank
elif 'SpikeData.h5' in files:
final_path = os.path.join(new_path, 'SpikeData.h5')
final_path = os.path.join(new_path, 'SpikeData.h5')
store = pd.HDFStore(final_path, 'r')
toreturn = {}
for n in store['neurons_id'].values:
toreturn[n] = nts.Ts(store['neuron_' + str(n)].index.values,
time_units='s')
shank = store['shanks'].values
store.close()
return toreturn, shank
else:
print("Couldn't find any SpikeData file in " + new_path)
print("If clu and res files are present in " + path +
", a SpikeData.h5 is going to be created")
# Creating /Analysis/ Folder here if not already present
if not os.path.exists(new_path): os.makedirs(new_path)
files = os.listdir(path)
clu_files = np.sort([f for f in files if '.clu.' in f and f[0] != '.'])
res_files = np.sort([f for f in files if '.res.' in f and f[0] != '.'])
clu1 = np.sort([int(f.split(".")[-1]) for f in clu_files])
clu2 = np.sort([int(f.split(".")[-1]) for f in res_files])
if len(clu_files) != len(res_files) or not (clu1 == clu2).any():
print("Not the same number of clu and res files in " + path +
"; Exiting ...")
sys.exit()
count = 0
neurons = {}
shank = []
for i in range(len(clu_files)):
clu = np.genfromtxt(os.path.join(path, clu_files[i]),
dtype=np.int32)[1:]
if np.max(clu) > 1:
res = np.genfromtxt(os.path.join(path, res_files[i]))
tmp = np.unique(clu).astype(int)
idx_clu = tmp[tmp > 1]
idx_neu = np.arange(count, count + len(idx_clu))
for j, n in zip(idx_clu, idx_neu):
neurons[n] = pd.Series(index=np.unique(res[clu == j]) / fs,
data=np.uint8(n),
dtype=np.uint8)
shank.append(int(clu_files[i].split(".")[-1]))
count += len(idx_clu)
# Saving SpikeData.h5
final_path = os.path.join(new_path, 'SpikeData.h5')
store = pd.HDFStore(final_path, 'w')
for n in neurons:
store.append('neuron_' + str(n), neurons[n])
store.append('neurons_id', pd.Series(np.array(list(neurons.keys()))))
store.append('shanks', pd.Series(shank))
store.close()
# Returning a dictionnary
toreturn = {}
for i in neurons:
toreturn[i] = nts.Ts(t=neurons[i].index.values, time_units='s')
return toreturn, shank
This script will introduce you to the basics of neuroseries It is the package used to handle spike times, epoch of wake/rem/sleep, etc It is build on pandas ''' import numpy as np import pandas as pd import neuroseries as nts from pylab import * # let's create fake data for example the time of 10 spikes between 0 and 15 s random_times = np.random.uniform(0, 15, 10) # let's sort them in ascending order of apparition random_times = np.sort(random_times) # we can include them in a neuroserie object called a Ts (Time series) my_spike = nts.Ts(random_times, time_units='s') # DON'T FORGET THE time_units otherwise it will consider you have spikes in microseconds # Observe your dataset my_spike # The first column indicates the timestamps in microseconds # The second column is full of NaN (Not A Number) because it's just time stamps # Let's try with spikes with milliseconds timestamps my_spike2 = nts.Ts(random_times, time_units='ms') # Observe the difference between the 2 my_spike my_spike2 # SO REMEMBER # ALWAYS CHECK THAT YOUR TIME UNITS ARE CORRECT! # If you have timestamps associated with a value for example 15 points of EEG during 15seconds my_eeg = np.sin(np.arange(0, 15))
def loadSpikeData(path, index=None, fs=20000):
"""
if the path contains a folder named /Analysis,
the script will look into it to load either
- SpikeData.mat saved from matlab
- SpikeData.h5 saved from this same script
if not, the res and clu file will be loaded
and an /Analysis folder will be created to save the data
Thus, the next loading of spike times will be faster
Notes :
If the frequency is not givne, it's assumed 20kH
Args:
path : string
Returns:
dict, array
"""
if not os.path.exists(path):
print("The path " + path + " doesn't exist; Exiting ...")
sys.exit()
new_path = os.path.join(path, 'Analysis/')
if os.path.exists(new_path):
new_path = os.path.join(path, 'Analysis/')
files = os.listdir(new_path)
if 'SpikeData.mat' in files:
spikedata = scipy.io.loadmat(new_path + 'SpikeData.mat')
shank = spikedata['shank'] - 1
if index is None:
shankIndex = np.arange(len(shank))
else:
shankIndex = np.where(shank == index)[0]
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(
spikedata['S'][0][0][0][i][0][0][0][1][0][0][2],
time_units='s')
a = spikes[0].as_units('s').index.values
if ((a[-1] - a[0]) / 60.) / 60. > 20.: # VERY BAD
spikes = {}
for i in shankIndex:
spikes[i] = nts.Ts(
spikedata['S'][0][0][0][i][0][0][0][1][0][0][2] *
0.0001,
time_units='s')
return spikes, shank
elif 'SpikeData.h5' in files:
final_path = os.path.join(new_path, 'SpikeData.h5')
spikes = pd.read_hdf(final_path, mode='r')
# Returning a dictionnary | can be changed to return a dataframe
toreturn = {}
for i, j in spikes:
toreturn[j] = nts.Ts(t=spikes[(i, j)].replace(
0, np.nan).dropna().index.values,
time_units='s')
shank = spikes.columns.get_level_values(0).values[:, np.newaxis]
return toreturn, shank
else:
print("Couldn't find any SpikeData file in " + new_path)
print("If clu and res files are present in " + path +
", a SpikeData.h5 is going to be created")
# Creating /Analysis/ Folder here if not already present
if not os.path.exists(new_path): os.makedirs(new_path)
files = os.listdir(path)
clu_files = np.sort([f for f in files if 'clu' in f and f[0] != '.'])
res_files = np.sort([f for f in files if 'res' in f and f[0] != '.'])
clu1 = np.sort([int(f.split(".")[-1]) for f in clu_files])
clu2 = np.sort([int(f.split(".")[-1]) for f in res_files])
if len(clu_files) != len(res_files) or not (clu1 == clu2).any():
print("Not the same number of clu and res files in " + path +
"; Exiting ...")
sys.exit()
count = 0
spikes = []
for i in range(len(clu_files)):
clu = np.genfromtxt(os.path.join(path, clu_files[i]),
dtype=np.int32)[1:]
if np.max(clu) > 1:
res = np.genfromtxt(os.path.join(path, res_files[i]))
tmp = np.unique(clu).astype(int)
idx_clu = tmp[tmp > 1]
idx_col = np.arange(count, count + len(idx_clu))
tmp = pd.DataFrame(index=np.unique(res) / fs,
columns=pd.MultiIndex.from_product([[i],
idx_col]),
data=0,
dtype=np.int32)
for j, k in zip(idx_clu, idx_col):
tmp.loc[res[clu == j] / fs, (i, k)] = k + 1
spikes.append(tmp)
count += len(idx_clu)
# tmp2 = pd.DataFrame(index=res[clu==j]/fs, data = k+1, ))
# spikes = pd.concat([spikes, tmp2], axis = 1)
spikes = pd.concat(spikes, axis=1)
spikes = spikes.fillna(0)
spikes = spikes.astype(np.int32)
# Saving SpikeData.h5
final_path = os.path.join(new_path, 'SpikeData.h5')
spikes.columns.set_names(['shank', 'neuron'], inplace=True)
spikes.to_hdf(final_path, key='spikes', mode='w')
# Returning a dictionnary
toreturn = {}
for i, j in spikes:
toreturn[j] = nts.Ts(t=spikes[(i, j)].replace(
0, np.nan).dropna().index.values,
time_units='s')
shank = spikes.columns.get_level_values(0).values[:, np.newaxis].flatten()
return toreturn, shank
os.mkdir(data_directory+'/plots')
path =rootDir + '/' + ID + '/' + session
#count number of sessions
ns = int([i for i in os.listdir(path) if os.path.isdir(path+'/'+i)==True][-1][-1:])
if ns == 1:
episodes=['wake']
else:
episodes = ['wake' if i==wakepos else 'sleep' for i in list(range(ns+1))]
spikes, shank = loadSpikeData(data_directory)
n_channels, fs, shank_to_channel = loadXML(data_directory)
position = loadPosition(data_directory, events, episodes, n_ttl_channels = 2, optitrack_ch = 0)
wake_ep = loadEpoch(data_directory, 'wake', episodes)
if "sleep" in episodes:
sleep_ep = loadEpoch(data_directory, 'sleep')
ttl_track, ttl_opto_start, ttl_opto_end = loadTTLPulse2(os.path.join(data_directory, session+'_0_analogin.dat'), 2)
ttl_track = nts.Ts(ttl_track.index.values, time_units = 's')
ttl_opto_start = nts.Ts(ttl_opto_start.index.values, time_units = 's')
ttl_opto_end = nts.Ts(ttl_opto_end.index.values, time_units = 's')
opto_ep = nts.IntervalSet(start = ttl_opto_start.index.values, end = ttl_opto_end.index.values)
stim_ep=manage.optoeps(ttl_opto_start, ttl_opto_end) #Load main stim epochs
neuron = 5
# whole duration= int(spikes[neuron].as_units('s').index[-1] - spikes[21].as_units('s').index[0])
bins = np.arange(0, 300000, 1)
plt.figure()
plt.hist(np.diff(spikes[neuron].index.values), bins)
bins = np.arange(0, 300000+10000, 10000)
'/Analysis/HDCells.mat')['hdCellStats'][:, -1]
hd_info_neuron = np.array([hd_info[n] for n in spikes.keys()])
spikes = {
k: spikes[k]
for k in np.where(hd_info_neuron == 0)[0] if k not in []
}
neurons = np.sort(list(spikes.keys()))
####################################################################################################################
# binning data
####################################################################################################################
allrates = {}
n_ex = 200
tmp = nts.Ts(rip_tsd.as_series().sample(
n_ex, replace=False).sort_index()).index.values
rip_tsd = pd.Series(index=tmp, data=np.nan)
# rip_tsd = rip_tsd.iloc[0:200]
bins_size = [200, 10, 100]
####################################################################################################################
# BIN SWR
####################################################################################################################
@jit(nopython=True)
def histo(spk, obins):
n = len(obins)
count = np.zeros(n)
for i in range(n):
count[i] = np.sum((spk > obins[i, 0]) * (spk < obins[i, 1]))
This script will introduce you to the basics of neuroseries It is the package used to handle spike time, epoch of wake/rem/sleep, etc It is build on pandas ''' import numpy as np import pandas as pd # the folder neuroseries should be in current directory # to check, you can type ls in ipython3 to list all the files and folders in your current directory import neuroseries as nts # let's create fake data for example the time of 10 spikes between 0 and 15 s my_spike = np.random.uniform(0, 15, 10) # let's sort them in ascending order of apparition my_spike = np.sort(my_spike) # we can include them in a neuroserie object called a Ts (Time series) my_spike = nts.Ts(my_spike, time_units='s') # DON'T FORGET THE time_units otherwise it will consider you have spikes in microseconds # Observe your dataset my_spike # The first column indicates the timestamps in microseconds # The second column is full of NaN (Not A Number) because it's just time stamps # IF you have timestamps associated with a value for example 15 points of EEG each second my_eeg = np.sin(np.arange(0, 15)) # You use a Tsd (Time series data) my_eeg = nts.Tsd(t=np.arange(15), d=my_eeg, time_units='s') # Observe your variable my_eeg # And how the software transform you timetamps in second in timestamps in microsecond # Now if you are using a fancy probe and recording for example 3 channel at the same times
subplot(211)
plot(angle1.mean(1), label = session)
legend()
title("Pas de zscore")
subplot(212)
plot(angle2.mean(1), label = session)
title("Zscored")
legend()
show()
sys.exit()
# UP/DOWN
up_tsd = nts.Ts(t = up_ep['start'].values)
d = np.vstack(rip_tsd.index.values) - up_tsd.index.values
d[d<0] = np.max(d)
idx = np.argmin(d, 1)
up_time = up_tsd.index.values[idx]
interval = rip_tsd.index.values - up_time
a = angle2.iloc[:,np.argsort(interval)]
groups = []
for idx in np.array_split(np.arange(a.shape[1]),3):
groups.append(a[idx].mean(1))
groups = pd.concat(groups, 1)
# let's imagine that the animal is moving his head clockwise at a constant speed angle = np.arange(0, 100, 0.1) # let's bring that between 0 and 2pi angle = angle % (2 * np.pi) # let's imagine the sampling rate is 100Hz for detecting the position of the animal # So we have a dt of dt = 1 / 100. # and a duration of duration = dt * len(angle) # let's put angle in a neuroseries tsd angle = nts.Tsd(t=np.arange(0, duration, dt), d=angle, time_units='s') # now let's imagine we have some spikes spikes = np.sort(np.random.uniform(0, duration, 100)) spikes = nts.Ts(spikes, time_units='s') # We can plot both angle and spikes together figure() plot(angle) plot(spikes.times(), np.zeros(len(spikes)), '|', markersize=10) show() #So the question is: What was the corresponding angular position when a spike was recorded # To do that, you use realign which basically takes the closed angle value in time from the spike angle_spike = angle.realign(spikes) # The order matters here! it's not spikes.realign(angle) # let's look at what it does figure() plot(angle) plot(spikes.times(), np.zeros(len(spikes)), '|', markersize=10)
spind_hpc_ep = nts.IntervalSet(tmp[:, 0], tmp[:, 1], time_units='ms')
spind_ep = spind_hpc_ep.intersect(spind_thl_ep).drop_short_intervals(0.0)
spind_thl_no_hpc = spind_thl_ep.set_diff(
spind_hpc_ep).drop_short_intervals(0.0)
spind_hpc_no_thl = spind_hpc_ep.set_diff(
spind_thl_ep).drop_short_intervals(0.0)
store = pd.HDFStore("../data/phase_spindles/" + session.split("/")[1] +
".lfp")
phase_hpc = nts.Tsd(store['phase_hpc_spindles'])
phase_thl = nts.Tsd(store['phase_thl_spindles'][0])
store.close()
spikes = {}
store_spike = pd.HDFStore("../data/spikes_thalamus/" +
session.split("/")[1] + ".spk")
for n in store_spike.keys():
spikes[int(n[1:])] = nts.Ts(store_spike[n])
store_spike.close()
##################################################################################################
# SPINDLES MODULATION
##################################################################################################
spind_mod1 = computePhaseModulation(phase_hpc, spikes, spind_hpc_ep)
spind_mod2 = computePhaseModulation(phase_thl, spikes, spind_thl_ep)
spind_mod3 = computePhaseModulation(phase_hpc, spikes, spind_ep)
spind_mod4 = computePhaseModulation(phase_thl, spikes, spind_ep)
spind_mod5 = computePhaseModulation(phase_thl, spikes, spind_thl_no_hpc)
spind_mod6 = computePhaseModulation(phase_hpc, spikes, spind_hpc_no_thl)
kappa = np.vstack([
spind_mod1[:, 2], spind_mod3[:, 2], spind_mod2[:, 2], spind_mod4[:, 2]
rip_ep,rip_tsd = loadRipples(data_directory+session)
rip_ep = sws_ep.intersect(rip_ep)
rip_tsd = rip_tsd.restrict(sws_ep)
hd_info = scipy.io.loadmat(data_directory+session+'/Analysis/HDCells.mat')['hdCellStats'][:,-1]
hd_info_neuron = np.array([hd_info[n] for n in spikes.keys()])
spikeshd = {k:spikes[k] for k in np.where(hd_info_neuron==1)[0] if k not in []}
spikesnohd = {k:spikes[k] for k in np.where(hd_info_neuron==0)[0] if k not in []}
hdneurons = np.sort(list(spikeshd.keys()))
nohdneurons = np.sort(list(spikesnohd.keys()))
# UP/DOWN
down_ep, up_ep = loadUpDown(data_directory+session)
up_tsd = nts.Ts(t = up_ep['start'].values)
bin_size = 10
nb_bins = 400
times = np.arange(0, bin_size*(nb_bins+1), bin_size) - (nb_bins*bin_size)/2
C = crossCorr(up_tsd.as_units('ms').index.values, rip_tsd.as_units('ms').index.values, bin_size, nb_bins)
C = pd.Series(index = times, data = C)
CC[session] = C
CC = pd.DataFrame.from_dict(CC)
a = CC.rolling(window = 40, win_type = 'gaussian', center = True, min_periods = 1).mean(std = 4.0)