from parser.CellData import CellData
from models.LIFACnoise import LifacNoiseModel
from stimuli.SinusoidalStepStimulus import SinusoidalStepStimulus
import numpy as np
import matplotlib.pyplot as plt
from warnings import warn
from my_util import helperFunctions as hF, functions as fu
from os.path import join, exists
import pickle
from sys import stderr
class FICurve:
def __init__(self, stimulus_values, save_dir=None, recalculate=False):
self.save_file_name = "fi_curve_values.pkl"
self.stimulus_values = stimulus_values
self.indices_f_baseline = []
self.f_baseline_frequencies = []
self.indices_f_inf = []
self.f_inf_frequencies = []
self.indices_f_zero = []
self.f_zero_frequencies = []
# increase, offset
self.f_inf_fit = []
# f_max, f_min, k, x_zero
self.f_zero_fit = []
if save_dir is None:
self.initialize()
else:
if recalculate:
self.initialize()
self.save_values(save_dir)
else:
if not self.load_values(save_dir):
self.initialize()
self.save_values(save_dir)
def initialize(self):
self.calculate_all_frequency_points()
self.f_inf_fit = hF.fit_clipped_line(self.stimulus_values, self.f_inf_frequencies)
self.f_zero_fit = hF.fit_boltzmann(self.stimulus_values, self.f_zero_frequencies)
def calculate_all_frequency_points(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_f_baseline_frequencies(self):
return self.f_baseline_frequencies
def get_f_inf_frequencies(self):
return self.f_inf_frequencies
def get_f_zero_frequencies(self):
return self.f_zero_frequencies
def get_f_inf_slope(self):
if len(self.f_inf_fit) > 0:
return self.f_inf_fit[0]
def get_f_zero_fit_slope_at_straight(self):
fit_vars = self.f_zero_fit
return fu.full_boltzmann_straight_slope(fit_vars[0], fit_vars[1], fit_vars[2], fit_vars[3])
def get_f_zero_fit_slope_at_stimulus_value(self, stimulus_value):
fit_vars = self.f_zero_fit
return fu.derivative_full_boltzmann(stimulus_value, fit_vars[0], fit_vars[1], fit_vars[2], fit_vars[3])
def get_f_inf_frequency_at_stimulus_value(self, stimulus_value):
return fu.clipped_line(stimulus_value, self.f_inf_fit[0], self.f_inf_fit[1])
def get_f_zero_and_f_inf_intersection(self):
x_values = np.arange(min(self.stimulus_values), max(self.stimulus_values), 0.0001)
fit_vars = self.f_zero_fit
f_zero = fu.full_boltzmann(x_values, fit_vars[0], fit_vars[1], fit_vars[2], fit_vars[3])
f_inf = fu.clipped_line(x_values, self.f_inf_fit[0], self.f_inf_fit[1])
intersection_indicies = np.argwhere(np.diff(np.sign(f_zero - f_inf))).flatten()
# print("fi-curve calc intersection:", intersection_indicies, x_values[intersection_indicies])
if len(intersection_indicies) > 1:
f_baseline = np.median(self.f_baseline_frequencies)
best_dist = np.inf
best_idx = -1
for idx in intersection_indicies:
dist = abs(fu.clipped_line(x_values[idx], self.f_inf_fit[0], self.f_inf_fit[1]) - f_baseline)
if dist < best_dist:
best_dist = dist
best_idx = idx
return x_values[best_idx]
elif len(intersection_indicies) == 0:
raise ValueError("No intersection found!")
else:
return x_values[intersection_indicies[0]]
def get_f_zero_fit_slope_at_f_inf_fit_intersection(self):
x = self.get_f_zero_and_f_inf_intersection()
fit_vars = self.f_zero_fit
return fu.derivative_full_boltzmann(x, fit_vars[0], fit_vars[1], fit_vars[2], fit_vars[3])
def get_mean_time_and_freq_traces(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_time_and_freq_traces(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_sampling_interval(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_delay(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_stimulus_start(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_stimulus_end(self):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def get_stimulus_duration(self):
return self.get_stimulus_end() - self.get_stimulus_start()
def plot_mean_frequency_curves(self, save_path=None):
time_traces, freq_traces = self.get_time_and_freq_traces()
mean_times, mean_freqs = self.get_mean_time_and_freq_traces()
for i, sv in enumerate(self.stimulus_values):
for j in range(len(time_traces[i])):
plt.plot(time_traces[i][j], freq_traces[i][j], color="gray", alpha=0.8)
plt.plot(mean_times[i], mean_freqs[i], color="black")
plt.xlabel("Time [s]")
plt.ylabel("Frequency [Hz]")
plt.title("Mean frequency at contrast {:.2f} ({:} trials)".format(sv, len(time_traces[i])))
if save_path is None:
plt.show()
else:
plt.savefig(save_path + "mean_frequency_contrast_{:.2f}.png".format(sv))
plt.close()
def plot_fi_curve(self, save_path=None):
min_x = min(self.stimulus_values)
max_x = max(self.stimulus_values)
step = (max_x - min_x) / 5000
x_values = np.arange(min_x, max_x, step)
plt.plot(self.stimulus_values, self.f_baseline_frequencies, color='blue', label='f_base')
plt.plot(self.stimulus_values, self.f_inf_frequencies, 'o', color='green', label='f_inf')
plt.plot(x_values, [fu.clipped_line(x, self.f_inf_fit[0], self.f_inf_fit[1]) for x in x_values],
color='darkgreen', label='f_inf_fit')
plt.plot(self.stimulus_values, self.f_zero_frequencies, 'o', color='orange', label='f_zero')
popt = self.f_zero_fit
plt.plot(x_values, [fu.full_boltzmann(x, popt[0], popt[1], popt[2], popt[3]) for x in x_values],
color='red', label='f_0_fit')
plt.legend()
plt.ylabel("Frequency [Hz]")
plt.xlabel("Stimulus value")
if save_path is None:
plt.show()
else:
plt.savefig(save_path + "fi_curve.png")
plt.close()
@staticmethod
def plot_fi_curve_comparision(data_fi_curve, model_fi_curve, save_path=None):
min_x = min(min(data_fi_curve.stimulus_values), min(model_fi_curve.stimulus_values))
max_x = max(max(data_fi_curve.stimulus_values), max(model_fi_curve.stimulus_values))
step = (max_x - min_x) / 5000
x_values = np.arange(min_x, max_x+step, step)
fig, axes = plt.subplots(1, 3, sharex="all", sharey='all', figsize=(15, 6))
# plot baseline
data_origin = (data_fi_curve, model_fi_curve)
f_base_color = ("blue", "deepskyblue")
f_inf_color = ("green", "limegreen")
f_zero_color = ("red", "orange")
for i in range(2):
axes[i].plot(data_origin[i].stimulus_values, data_origin[i].get_f_baseline_frequencies(),
color=f_base_color[i], label='f_base')
axes[i].plot(data_origin[i].stimulus_values, data_origin[i].get_f_inf_frequencies(),
'o', color=f_inf_color[i], label='f_inf')
y_values = [fu.clipped_line(x, data_origin[i].f_inf_fit[0], data_origin[i].f_inf_fit[1]) for x in x_values]
axes[i].plot(x_values, y_values, color=f_inf_color[i], label='f_inf_fit')
axes[i].plot(data_origin[i].stimulus_values, data_origin[i].get_f_zero_frequencies(),
'o', color=f_zero_color[i], label='f_zero')
popt = data_origin[i].f_zero_fit
axes[i].plot(x_values, [fu.full_boltzmann(x, popt[0], popt[1], popt[2], popt[3]) for x in x_values],
color=f_zero_color[i], label='f_0_fit')
axes[i].set_xlabel("Stimulus value - contrast")
axes[i].legend()
axes[0].set_title("cell")
axes[0].set_ylabel("Frequency [Hz]")
axes[1].set_title("model")
median_baseline = np.median(data_fi_curve.get_f_baseline_frequencies())
axes[2].plot((min_x, max_x), (median_baseline, median_baseline), color=f_base_color[0], label="cell med base")
axes[2].plot(model_fi_curve.stimulus_values, model_fi_curve.get_f_baseline_frequencies(),
'o', color=f_base_color[1], label='model base')
y_values = [fu.clipped_line(x, data_fi_curve.f_inf_fit[0], data_fi_curve.f_inf_fit[1]) for x in x_values]
axes[2].plot(x_values, y_values, color=f_inf_color[0], label='f_inf_fit cell')
axes[2].plot(model_fi_curve.stimulus_values, model_fi_curve.get_f_inf_frequencies(),
'o', color=f_inf_color[1], label='f_inf model')
popt = data_fi_curve.f_zero_fit
axes[2].plot(x_values, [fu.full_boltzmann(x, popt[0], popt[1], popt[2], popt[3]) for x in x_values],
color=f_zero_color[0], label='f_0_fit cell')
axes[2].plot(model_fi_curve.stimulus_values, model_fi_curve.get_f_zero_frequencies(),
'o', color=f_zero_color[1], label='f_zero model')
axes[2].set_title("cell model comparision")
axes[2].set_xlabel("Stimulus value - contrast")
axes[2].legend()
if save_path is None:
plt.show()
else:
plt.savefig(save_path + "fi_curve_comparision.png")
plt.close()
def plot_f_point_detections(self, save_path=None):
raise NotImplementedError("NOT YET OVERRIDDEN FROM ABSTRACT CLASS")
def save_values(self, save_directory):
values = {}
values["stimulus_values"] = self.stimulus_values
values["f_baseline_frequencies"] = self.f_baseline_frequencies
values["f_inf_frequencies"] = self.f_inf_frequencies
values["f_zero_frequencies"] = self.f_zero_frequencies
values["f_inf_fit"] = self.f_inf_fit
values["f_zero_fit"] = self.f_zero_fit
with open(join(save_directory, self.save_file_name), "wb") as file:
pickle.dump(values, file)
print("Fi-Curve: Values saved!")
def load_values(self, save_directory):
file_path = join(save_directory, self.save_file_name)
if not exists(file_path):
print("Fi-Curve: No file to load")
return False
file = open(file_path, "rb")
values = pickle.load(file)
if set(values["stimulus_values"]) != set(self.stimulus_values):
stderr.write("Fi-Curve:load_values() - Given stimulus values are different to the loaded ones!:\n "
"given: {}\n loaded: {}\n".format(str(self.stimulus_values), str(values["stimulus_values"])))
self.stimulus_values = values["stimulus_values"]
self.f_baseline_frequencies = values["f_baseline_frequencies"]
self.f_inf_frequencies = values["f_inf_frequencies"]
self.f_zero_frequencies = values["f_zero_frequencies"]
self.f_inf_fit = values["f_inf_fit"]
self.f_zero_fit = values["f_zero_fit"]
print("Fi-Curve: Values loaded!")
return True
class FICurveCellData(FICurve):
def __init__(self, cell_data: CellData, stimulus_values, save_dir=None, recalculate=False):
self.cell_data = cell_data
super().__init__(stimulus_values, save_dir, recalculate)
def calculate_all_frequency_points(self):
mean_frequencies = self.cell_data.get_mean_fi_curve_isi_frequencies()
time_axes = self.cell_data.get_time_axes_fi_curve_mean_frequencies()
stimulus_start = self.cell_data.get_stimulus_start()
stimulus_duration = self.cell_data.get_stimulus_duration()
sampling_interval = self.cell_data.get_sampling_interval()
if len(mean_frequencies) == 0:
warn("FICurve:all_calculate_frequency_points(): mean_frequencies is empty.\n"
"Was all_calculate_mean_isi_frequencies already called?")
for i in range(len(mean_frequencies)):
if time_axes[i][0] > self.cell_data.get_stimulus_start():
raise ValueError("TODO: Deal with to strongly cut frequency traces in cell data! ")
# self.f_zero_frequencies.append(-1)
# self.f_baseline_frequencies.append(-1)
# self.f_inf_frequencies.append(-1)
# continue
f_zero, f_zero_idx = hF.detect_f_zero_in_frequency_trace(time_axes[i], mean_frequencies[i],
stimulus_start, sampling_interval)
self.f_zero_frequencies.append(f_zero)
self.indices_f_zero.append(f_zero_idx)
f_baseline, f_base_idx = hF.detect_f_baseline_in_freq_trace(time_axes[i], mean_frequencies[i],
stimulus_start, sampling_interval)
self.f_baseline_frequencies.append(f_baseline)
self.indices_f_baseline.append(f_base_idx)
f_infinity, f_inf_idx = hF.detect_f_infinity_in_freq_trace(time_axes[i], mean_frequencies[i],
stimulus_start, stimulus_duration, sampling_interval)
self.f_inf_frequencies.append(f_infinity)
self.indices_f_inf.append(f_inf_idx)
def get_mean_time_and_freq_traces(self):
return self.cell_data.get_time_axes_fi_curve_mean_frequencies(), self.cell_data.get_mean_fi_curve_isi_frequencies()
def get_time_and_freq_traces(self):
spiketimes = self.cell_data.get_fi_spiketimes()
time_traces = []
freq_traces = []
for i in range(len(spiketimes)):
trial_time_traces = []
trial_freq_traces = []
for j in range(len(spiketimes[i])):
time, isi_freq = hF.calculate_time_and_frequency_trace(spiketimes[i][j], self.cell_data.get_sampling_interval())
trial_freq_traces.append(isi_freq)
trial_time_traces.append(time)
time_traces.append(trial_time_traces)
freq_traces.append(trial_freq_traces)
return time_traces, freq_traces
def get_sampling_interval(self):
return self.cell_data.get_sampling_interval()
def get_delay(self):
return self.cell_data.get_delay()
def get_stimulus_start(self):
return self.cell_data.get_stimulus_start()
def get_stimulus_end(self):
return self.cell_data.get_stimulus_end()
def get_f_zero_inverse_at_frequency(self, frequency):
# UNUSED
b_vars = self.f_zero_fit
return fu.inverse_full_boltzmann(frequency, b_vars[0], b_vars[1], b_vars[2], b_vars[3])
def get_f_infinity_frequency_at_stimulus_value(self, stimulus_value):
# UNUSED
infty_vars = self.f_inf_fit
return fu.clipped_line(stimulus_value, infty_vars[0], infty_vars[1])
def plot_f_point_detections(self, save_path=None):
mean_frequencies = np.array(self.cell_data.get_mean_fi_curve_isi_frequencies())
time_axes = self.cell_data.get_time_axes_fi_curve_mean_frequencies()
sampling_interval = self.cell_data.get_sampling_interval()
stim_start = self.cell_data.get_stimulus_start()
stim_duration = self.cell_data.get_stimulus_duration()
for i, c in enumerate(self.stimulus_values):
time = time_axes[i]
frequency = mean_frequencies[i]
if len(time) == 0 or min(time) > stim_start \
or max(time) < stim_start + stim_duration:
continue
fig, ax = plt.subplots(1, 1, figsize=(8, 8))
ax.set_title("Stimulus value: {:.2f}".format(c))
ax.plot(time, frequency)
start_idx, end_idx = hF.time_window_detect_f_baseline(time[0], stim_start, sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_baseline_frequencies[i], self.f_baseline_frequencies[i]),
label="f_base", color="deepskyblue")
start_idx, end_idx = hF.time_window_detect_f_infinity(time[0], stim_start, stim_duration,
sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_inf_frequencies[i], self.f_inf_frequencies[i]),
label="f_inf", color="limegreen")
start_idx, end_idx = hF.time_window_detect_f_zero(time[0], stim_start, sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_zero_frequencies[i], self.f_zero_frequencies[i]),
label="f_zero", color="orange")
plt.legend()
if save_path is not None:
plt.savefig(save_path + "/detections_contrast_{:.2f}.png".format(c))
else:
plt.show()
plt.close()
class FICurveModel(FICurve):
stim_duration = 0.5
stim_start = 0.5
total_simulation_time = stim_duration + 2 * stim_start
def __init__(self, model, stimulus_values, eod_frequency, trials=5):
self.eod_frequency = eod_frequency
self.model = model
self.trials = trials
self.spiketimes_array = np.zeros((len(stimulus_values), trials), dtype=list)
self.mean_frequency_traces = []
self.mean_time_traces = []
self.set_model_adaption_to_baseline()
super().__init__(stimulus_values)
def set_model_adaption_to_baseline(self):
stimulus = SinusoidalStepStimulus(self.eod_frequency, 0, 0, 0)
self.model.simulate(stimulus, 1)
adaption = self.model.get_adaption_trace()
self.model.set_variable("a_zero", adaption[-1])
# print("FiCurve: model a_zero set to", adaption[-1])
def calculate_all_frequency_points(self):
sampling_interval = self.model.get_sampling_interval()
self.f_inf_frequencies = []
self.f_zero_frequencies = []
self.f_baseline_frequencies = []
for i, c in enumerate(self.stimulus_values):
stimulus = SinusoidalStepStimulus(self.eod_frequency, c, self.stim_start, self.stim_duration)
frequency_traces = []
time_traces = []
for j in range(self.trials):
_, spiketimes = self.model.simulate(stimulus, self.total_simulation_time)
self.spiketimes_array[i, j] = spiketimes
trial_time, trial_frequency = hF.calculate_time_and_frequency_trace(spiketimes, sampling_interval)
frequency_traces.append(trial_frequency)
time_traces.append(trial_time)
time, frequency = hF.calculate_mean_of_frequency_traces(time_traces, frequency_traces, sampling_interval)
self.mean_frequency_traces.append(frequency)
self.mean_time_traces.append(time)
if len(time) == 0 or min(time) > self.stim_start \
or max(time) < self.stim_start + self.stim_duration:
# print("Too few spikes to calculate f_inf, f_0 and f_base")
self.f_inf_frequencies.append(0)
self.f_zero_frequencies.append(0)
self.f_baseline_frequencies.append(0)
continue
f_inf, f_inf_idx = hF.detect_f_infinity_in_freq_trace(time, frequency, self.stim_start, self.stim_duration, sampling_interval)
self.f_inf_frequencies.append(f_inf)
self.indices_f_inf.append(f_inf_idx)
f_zero, f_zero_idx = hF.detect_f_zero_in_frequency_trace(time, frequency, self.stim_start, sampling_interval)
self.f_zero_frequencies.append(f_zero)
self.indices_f_zero.append(f_zero_idx)
f_baseline, f_base_idx = hF.detect_f_baseline_in_freq_trace(time, frequency, self.stim_start, sampling_interval)
self.f_baseline_frequencies.append(f_baseline)
self.indices_f_baseline.append(f_base_idx)
def get_mean_time_and_freq_traces(self):
return self.mean_time_traces, self.mean_frequency_traces
def get_sampling_interval(self):
return self.model.get_sampling_interval()
def get_delay(self):
return 0
def get_stimulus_start(self):
return self.stim_start
def get_stimulus_end(self):
return self.stim_start + self.stim_duration
def get_time_and_freq_traces(self):
time_traces = []
freq_traces = []
for v in range(len(self.stimulus_values)):
times_for_value = []
freqs_for_value = []
for s in self.spiketimes_array[v]:
t, f = hF.calculate_time_and_frequency_trace(s, self.model.get_sampling_interval())
times_for_value.append(t)
freqs_for_value.append(f)
time_traces.append(times_for_value)
freq_traces.append(freqs_for_value)
return time_traces, freq_traces
def plot_f_point_detections(self, save_path=None):
sampling_interval = self.model.get_sampling_interval()
for i, c in enumerate(self.stimulus_values):
time = self.mean_time_traces[i]
frequency = self.mean_frequency_traces[i]
if len(time) == 0 or min(time) > self.stim_start \
or max(time) < self.stim_start + self.stim_duration:
continue
fig, ax = plt.subplots(1, 1, figsize=(8, 8))
ax.plot(time, frequency)
start_idx, end_idx = hF.time_window_detect_f_baseline(time[0], self.stim_start, sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_baseline_frequencies[i], self.f_baseline_frequencies[i]),
label="f_base", color="deepskyblue")
start_idx, end_idx = hF.time_window_detect_f_infinity(time[0], self.stim_start, self.stim_duration, sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_inf_frequencies[i], self.f_inf_frequencies[i]),
label="f_inf", color="limegreen")
start_idx, end_idx = hF.time_window_detect_f_zero(time[0], self.stim_start, sampling_interval)
ax.plot((time[start_idx], time[end_idx]), (self.f_zero_frequencies[i], self.f_zero_frequencies[i]),
label="f_zero", color="orange")
plt.legend()
if save_path is not None:
plt.savefig(save_path + "/detections_contrast_{:.2f}.png".format(c))
else:
plt.show()
plt.close()
def get_fi_curve_class(data, stimulus_values, eod_freq=None, trials=5, save_dir=None, recalculate=False) -> FICurve:
if isinstance(data, CellData):
return FICurveCellData(data, stimulus_values, save_dir, recalculate)
if isinstance(data, LifacNoiseModel):
if eod_freq is None:
raise ValueError("The FiCurveModel needs the eod variable to work")
return FICurveModel(data, stimulus_values, eod_freq, trials=trials)
raise ValueError("Unknown type: Cannot find corresponding Baseline class. Data was type:" + str(type(data)))