def preprocess_short_square_sweeps(data_set,
sweep_numbers,
extra_dur=0.2,
spike_window=0.05):
if len(sweep_numbers) == 0:
raise er.FeatureError(
"No short square sweeps available for feature extraction")
good_sweep_numbers, ssq_start, ssq_end = validate_sweeps(
data_set, sweep_numbers, extra_dur=extra_dur)
if len(good_sweep_numbers) == 0:
raise er.FeatureError(
"No short square sweeps were long enough or did not end early")
ssq_sweeps = data_set.sweep_set(good_sweep_numbers)
ssq_spx, ssq_spfx = dsf.extractors_for_sweeps(
ssq_sweeps,
est_window=[ssq_start, ssq_start + 0.001],
start=ssq_start,
end=ssq_end + spike_window,
reject_at_stim_start_interval=0.0002,
**dsf.detection_parameters(data_set.SHORT_SQUARE))
ssq_an = spa.ShortSquareAnalysis(ssq_spx, ssq_spfx)
ssq_features = ssq_an.analyze(ssq_sweeps)
return ssq_sweeps, ssq_features
def extract_features(data_set, ramp_sweep_numbers, ssq_sweep_numbers, lsq_sweep_numbers,
amp_interval=20, max_above_rheo=100):
features = {}
# RAMP FEATURES -----------------
if len(ramp_sweep_numbers) > 0:
ramp_sweeps = data_set.sweep_set(ramp_sweep_numbers)
ramp_start, ramp_dur, _, _, _ = stf.get_stim_characteristics(ramp_sweeps.sweeps[0].i, ramp_sweeps.sweeps[0].t)
ramp_spx, ramp_spfx = dsf.extractors_for_sweeps(ramp_sweeps,
start = ramp_start,
**dsf.detection_parameters(data_set.RAMP))
ramp_an = spa.RampAnalysis(ramp_spx, ramp_spfx)
basic_ramp_features = ramp_an.analyze(ramp_sweeps)
first_spike_ramp_features = first_spike_ramp(ramp_an)
features.update(first_spike_ramp_features)
# SHORT SQUARE FEATURES -----------------
if len(ssq_sweep_numbers) > 0:
ssq_sweeps = data_set.sweep_set(ssq_sweep_numbers)
ssq_start, ssq_dur, _, _, _ = stf.get_stim_characteristics(ssq_sweeps.sweeps[0].i, ssq_sweeps.sweeps[0].t)
ssq_spx, ssq_spfx = dsf.extractors_for_sweeps(ssq_sweeps,
est_window = [ssq_start, ssq_start+0.001],
**dsf.detection_parameters(data_set.SHORT_SQUARE))
ssq_an = spa.ShortSquareAnalysis(ssq_spx, ssq_spfx)
basic_ssq_features = ssq_an.analyze(ssq_sweeps)
first_spike_ssq_features = first_spike_ssq(ssq_an)
first_spike_ssq_features["short_square_current"] = basic_ssq_features["stimulus_amplitude"]
features.update(first_spike_ssq_features)
# LONG SQUARE SUBTHRESHOLD FEATURES -----------------
if len(lsq_sweep_numbers) > 0:
check_lsq_sweeps = data_set.sweep_set(lsq_sweep_numbers)
lsq_start, lsq_dur, _, _, _ = stf.get_stim_characteristics(check_lsq_sweeps.sweeps[0].i, check_lsq_sweeps.sweeps[0].t)
# Check that all sweeps are long enough and not ended early
extra_dur = 0.2
good_lsq_sweep_numbers = [n for n, s in zip(lsq_sweep_numbers, check_lsq_sweeps.sweeps)
if s.t[-1] >= lsq_start + lsq_dur + extra_dur and not np.all(s.v[tsu.find_time_index(s.t, lsq_start + lsq_dur)-100:tsu.find_time_index(s.t, lsq_start + lsq_dur)] == 0)]
lsq_sweeps = data_set.sweep_set(good_lsq_sweep_numbers)
lsq_spx, lsq_spfx = dsf.extractors_for_sweeps(lsq_sweeps,
start = lsq_start,
end = lsq_start + lsq_dur,
**dsf.detection_parameters(data_set.LONG_SQUARE))
lsq_an = spa.LongSquareAnalysis(lsq_spx, lsq_spfx, subthresh_min_amp=-100.)
basic_lsq_features = lsq_an.analyze(lsq_sweeps)
features.update({
"input_resistance": basic_lsq_features["input_resistance"],
"tau": basic_lsq_features["tau"],
"v_baseline": basic_lsq_features["v_baseline"],
"sag_nearest_minus_100": basic_lsq_features["sag"],
"sag_measured_at": basic_lsq_features["vm_for_sag"],
"rheobase_i": int(basic_lsq_features["rheobase_i"]),
"fi_linear_fit_slope": basic_lsq_features["fi_fit_slope"],
})
# TODO (maybe): port sag_from_ri code over
# Identify suprathreshold set for analysis
sweep_table = basic_lsq_features["spiking_sweeps"]
mask_supra = sweep_table["stim_amp"] >= basic_lsq_features["rheobase_i"]
sweep_indexes = fv._consolidated_long_square_indexes(sweep_table.loc[mask_supra, :])
amps = np.rint(sweep_table.loc[sweep_indexes, "stim_amp"].values - basic_lsq_features["rheobase_i"])
spike_data = np.array(basic_lsq_features["spikes_set"])
for amp, swp_ind in zip(amps, sweep_indexes):
if (amp % amp_interval != 0) or (amp > max_above_rheo) or (amp < 0):
continue
amp_label = int(amp / amp_interval)
first_spike_lsq_sweep_features = first_spike_lsq(spike_data[swp_ind])
features.update({"ap_1_{:s}_{:d}_long_square".format(f, amp_label): v
for f, v in first_spike_lsq_sweep_features.items()})
mean_spike_lsq_sweep_features = mean_spike_lsq(spike_data[swp_ind])
features.update({"ap_mean_{:s}_{:d}_long_square".format(f, amp_label): v
for f, v in mean_spike_lsq_sweep_features.items()})
sweep_feature_list = [
"first_isi",
"avg_rate",
"isi_cv",
"latency",
"median_isi",
"adapt",
]
features.update({"{:s}_{:d}_long_square".format(f, amp_label): sweep_table.at[swp_ind, f]
for f in sweep_feature_list})
features["stimulus_amplitude_{:d}_long_square".format(amp_label)] = int(amp + basic_lsq_features["rheobase_i"])
rates = sweep_table.loc[sweep_indexes, "avg_rate"].values
features.update(fi_curve_fit(amps, rates))
return features
