def find_extra_spikes(data, fast_data, fs): #, fast_freq = 750.):
""" finds spikes in the extracellular data, of fs sampling frequency
it will first highpass filter the data above vfast_freq to detect the spikes
"""
plot_it = False
allow_error = 0.2 # ms
alerr_pts = ms2pts(allow_error,fs)
# find spikes
spike_data = fast_data * (-1)
#spike_data[spike_data < 0] = 0
thres = np.std(spike_data)*4 # use thres as 6 * standard deviation
waves = dw.find_above(spike_data, thres)
starts, ends = dw.find_startend(waves) # finds all the peaks above threshold
mins, min_idxs = dw.max_waves(data*(-1), starts-int(alerr_pts), ends+int(alerr_pts)) # finds mins of each wave
amplitudes = np.mean(fast_data) - fast_data[min_idxs]
#import pdb; pdb.set_trace()
if plot_it:
scale = 'ms'
t = dat.get_timeline(data, fs, scale)
py.plot(t, data)
py.plot(t, spike_data)
py.plot(t[min_idxs], spike_data[min_idxs], 'o')
for m in range(len(min_idxs)):
py.text(t[min_idxs[m]], spike_data[min_idxs[m]], str(amplitudes[m]))
py.xlim([0,t[-1]])
py.show()
return amplitudes, min_idxs
def find_halfampl(data, data_stat, data_smooth, fs, min_idxs, rang = [1, 3]): #, fast_freq = 500., smooth_freq = [500, 2000]):
""" detects values such as:
a valley to peak
b half valley width
data filtered to fast_freq """
# cut the dataparts to given range
waves, min_idxs2 = cut_spikes(data_stat, fs, min_idxs) # fast data
waves_smooth, min_idxs2 = cut_spikes(data_smooth, fs, min_idxs) # slow data
norm_waves, min_idxs2 = cut_spikes(data, fs, min_idxs) # slow data
# initialize variables to find
half_bs, half_as, ampls, relative_ampl = [], [], [], []
As, Bs, norm_factor = [], [], []
peak = ms2pts(rang[0],fs) # place is always peak (in pts)
# go through every possible wave (spike)
for i in range(len(waves)):
if len(waves[i]) > 0:
# finding the maxs from left and right
#mleft, mright = max_leftright(waves[i], peak)
mleft_smooth, mright_smooth = max_leftright(waves_smooth[i], peak)
half_a = mright_smooth - peak
# finding the end of the spike
#temp, spadek = max_leftright(waves_smooth[i] * (-1), mright_smooth)
# get the amplitude of the spike before normalization (calculated on the original data)
maxRight = norm_waves[i][mright_smooth]
maxLeft = norm_waves[i][mleft_smooth]
top = max(maxRight, maxLeft)
temp_ampl = top - norm_waves[i][peak]
# if temp_ampl < 0:
# t = dat.get_timeline(waves_smooth[i], fs, 'ms')
# py.figure()
# py.plot(t, norm_waves[i], 'r')
# py.plot(t, waves[i], 'g')
# #py.plot(t, norm_waves2, 'b')
# #
# #py.plot(t[mleft], norm_waves[i][mleft], 'ob')
# #py.plot(t[mright], norm_waves[i][mright], 'ob')
# py.plot(t[mleft_smooth], norm_waves[i][mleft_smooth], 'og')
# py.plot(t[mright_smooth], norm_waves[i][mright_smooth], 'og')
# #y.plot(t[find_start], norm_waves[i][find_start], '<r')
#
# #py.vlines(t[peak], norm_waves[i][peak], norm_waves[i][peak]+temp_ampl)
# py.show()
# and relative amplitude: dist(maxleft to maxright)/dist(max_right to peak)
dist1 = norm_waves[i][mright_smooth] - norm_waves[i][mleft_smooth]
dist2 = np.abs(norm_waves[i][mleft_smooth] - norm_waves[i][peak])
temp_rel_amp = dist1/dist2
# normalize
# move the waves to the baseline
based = norm_waves[i][mleft_smooth]
norm_waves[i] = norm_waves[i] - based
max_pt = norm_waves[i][mleft_smooth]
min_pt = norm_waves[i][peak]
temp_norm_factor = np.abs(norm_waves[i][peak])
if min_pt == 0 or temp_ampl <= 0 or temp_norm_factor < 0: # or min_pt_smooth == 0:
#print 'appending 0 because amplitude is equal 0'
half_bs.append(0)
half_as.append(0)
#half_cs.append(0)
ampls.append(0)
relative_ampl.append(0)
As.append(0)
Bs.append(0)
norm_factor.append(0)
continue
else:
norm_waves[i] = norm_waves[i]/temp_norm_factor
# calculate half height as half between base and min (peak of the spike)
min_pt = -1.
half_pt = np.mean([max_pt, min_pt])
# find which part belongs to the spike below the half hight and calculate the width
all_below = np.zeros(np.size(norm_waves[i]))
all_below[waves[i] > half_pt] = 1
# calculate half_c (detection of D is not perfect so better not to use)
# C_c = norm_waves[i][mright_smooth]
# D_c = norm_waves[i][spadek]
# half_hC = (C_c + D_c)/2.0
#
# all_below_c = np.ones(np.size(waves_smooth[i]))
# all_below_c[waves_smooth[i] > half_hC] = 0
# t = dat.get_timeline(waves_smooth[i], fs, 'ms')
# py.figure()
# py.plot(t, norm_waves[i], 'r')
# py.plot(t, waves[i], 'g')
# py.plot(t, norm_waves2, 'b')
##
# #py.plot(t[mleft], norm_waves[i][mleft], 'ob')
# #py.plot(t[mright], norm_waves[i][mright], 'ob')
# py.plot(t[mleft_smooth], norm_waves[i][mleft_smooth], 'og')
# py.plot(t[mright_smooth], norm_waves[i][mright_smooth], 'og')
#y.plot(t[find_start], norm_waves[i][find_start], '<r')
#py.vlines(t[peak], norm_waves[i][peak], norm_waves[i][peak]+temp_ampl)
#py.show()
try:
# find_start_c = np.nonzero(all_below_c[:mright_smooth])[0][-1]
#print len(np.nonzero(all_below[:peak])[0])
if len(np.nonzero(all_below[:peak])[0]) == 0:
find_start = mleft_smooth
else:
find_start = np.nonzero(all_below[:peak])[0][-1]
if find_start < mleft_smooth:
find_start = mleft_smooth
find_end = np.nonzero(all_below[peak:])[0][0]+peak
# find_end_c = np.nonzero(all_below_c[mright_smooth:])[0][0]+mright_smooth
# if find_end > mright:
# find_end = mright
#
except IndexError:
#print 'appending 0 because of half widths'
half_bs.append(0)
half_as.append(0)
relative_ampl.append(0)
ampls.append(0)
As.append(0)
Bs.append(0)
norm_factor.append(0)
# continue
half = find_end - find_start
half_bs.append(half)
half_as.append(half_a)
ampls.append(temp_ampl)
relative_ampl.append(temp_rel_amp)
pk = ms2pts(rang[0], fs)
As.append(mleft_smooth - pk)
Bs.append(mright_smooth - pk)
norm_factor.append(temp_norm_factor)
t = dat.get_timeline(waves[i], fs, 'ms')
else:
#print 'appending 0 because wave is too short'
half_bs.append(0)
half_as.append(0)
#half_cs.append(0)
ampls.append(0)
relative_ampl.append(0)
As.append(0)
Bs.append(0)
norm_factor.append(0)
return half_as, half_bs, ampls, As, Bs, norm_factor
