def __str__(self):
out = [super().__str__()]
out.append('Analysis Directory: %s' % self.analysis_dir)
out.append('Recording Directories :')
out.append(pt.print_dict(self.rec_labels, tabs=1))
out.append('\nTaste Mapping :')
out.append(pt.print_dict(self.taste_map, tabs=1))
out.append('\nElectrode Mapping\n----------------')
out.append(pt.print_dataframe(self.electrode_mapping))
if hasattr(self, 'held_units'):
out.append('\nHeld Units :')
out.append(pt.print_dataframe(
self.held_units.drop(columns=['J3'])))
return '\n'.join(out)
def confirm_parameter_dict(params, prompt, shell=False):
'''Shows user a dictionary and asks them to confirm that the values are
correct. If not they have an option to edit the dict.
Parameters
----------
params: dict
values in dict can be int, float, str, bool, list, dict or None
prompt: str
prompt to show user
shell : bool (optional)
True to use command line interface
False (default) for GUI
Returns
-------
dict
lists are returned as lists of str so other types m ust be cast manually
by user
'''
prompt = ('----------\n%s\n----------\n%s\nAre these parameters good?' %
(prompt, pt.print_dict(params)))
q = ask_user(prompt, choices=['Yes', 'Edit', 'Cancel'], shell=shell)
if q == 2:
return None
elif q == 0:
return params
else:
new_params = fill_dict(params, 'Enter new values:', shell=shell)
return new_params
def make_unit_arrays(self):
'''Make spike arrays for each unit and store in hdf5 store
'''
params = self.spike_array_params
print('Generating unit arrays with parameters:\n----------')
print(pt.print_dict(params, tabs=1))
ss.make_spike_arrays(self.h5_file, params)
self.process_status['make_unit_arrays'] = True
self.save()
def read_rec_info(file_dir, shell=True):
'''Reads the info.rhd file to get relevant parameters.
Parameters
----------
file_dir : str, path to recording directory
Returns
-------
dict, necessary analysis info from info.rhd
fields: amplifier_sampling_rate, dig_in_sampling_rate, notch_filter,
ports (list, corresponds to channels), channels (list)
Throws
------
FileNotFoundError : if info.rhd is not in file_dir
'''
info_file = os.path.join(file_dir, 'info.rhd')
if not os.path.isfile(info_file):
raise FileNotFoundError('info.rhd file not found in %s' % file_dir)
out = {}
print('Reading info.rhd file...')
try:
info = load_intan_rhd_format.read_data(info_file)
except Exception as e:
# TODO: Have a way to manually input settings
info = None
userIO.tell_user('%s was unable to be read. May be corrupted or '
'recording may have been interrupted' % info_file,
shell=True)
raise e
freq_params = info['frequency_parameters']
notch_freq = freq_params['notch_filter_frequency']
amp_fs = freq_params['amplifier_sample_rate']
dig_in_fs = freq_params['board_dig_in_sample_rate']
out = {
'amplifier_sampling_rate': amp_fs,
'dig_in_sampling_rate': dig_in_fs,
'notch_filter': notch_freq
}
amp_ch = info['amplifier_channels']
ports = [x['port_prefix'] for x in amp_ch]
channels = [x['native_order'] for x in amp_ch]
out['ports'] = ports
out['channels'] = channels
out['num_channels'] = len(channels)
if info.get('board_dig_in_channels'):
dig_in = info['board_dig_in_channels']
din = [x['native_order'] for x in dig_in]
out['dig_in'] = din
if info.get('board_dig_out_channels'):
dig_out = info['board_dig_out_channels']
dout = [x['native_order'] for x in dig_out]
out['dig_out'] = dout
out['file_type'] = get_recording_filetype(file_dir)
print('\nRecording Info\n--------------\n')
print(pt.print_dict(out))
return out
def cluster_spikes(self,
data_quality=None,
multi_process=True,
n_cores=None,
custom_params=None,
umap=False):
'''Write clustering parameters to file and
Run blech_process on each electrode using GNU parallel
Parameters
----------
data_quality : {'clean', 'noisy', None (default)}
set if you want to change the data quality parameters for cutoff
and spike detection before running clustering. These parameters are
automatically set as "clean" during initial parameter setup
accept_params : bool, False (default)
set to True in order to skip popup confirmation of parameters when
running
'''
clustering_params = None
if custom_params:
clustering_params = custom_params
elif data_quality:
tmp = dio.params.load_params('clustering_params',
self.root_dir,
default_keyword=data_quality)
if tmp:
clustering_params = tmp
else:
raise ValueError('%s is not a valid data_quality preset. Must '
'be "clean" or "noisy" or None.')
# Get electrodes, throw out 'dead' electrodes
em = self.electrode_mapping
if 'dead' in em.columns:
electrodes = em.Electrode[em['dead'] == False].tolist()
else:
electrodes = em.Electrode.tolist()
# Setup progress bar
pbar = tqdm(total=len(electrodes))
def update_pbar(ans):
pbar.update()
# get clustering params
rec_dirs = list(self.rec_labels.values())
if clustering_params is None:
dat = load_dataset(rec_dirs[0])
clustering_params = dat.clustering_params.copy()
print('\nRunning Blech Clust\n-------------------')
print('Parameters\n%s' % pt.print_dict(clustering_params))
# Write clustering params to recording directories & check for spike detection
spike_detect = True
for rd in rec_dirs:
dat = load_dataset(rd)
if dat.process_status['spike_detection'] == False:
raise FileNotFoundError(
'Spike detection has not been run on %s' % rd)
dat.clustering_params = clustering_params
wt.write_params_to_json('clustering_params', rd, clustering_params)
# dat.save()
# Run clustering
if not umap:
clust_objs = [
bclust.BlechClust(rec_dirs, x, params=clustering_params)
for x in electrodes
]
else:
clust_objs = [
bclust.BlechClust(rec_dirs,
x,
params=clustering_params,
data_transform=bclust.UMAP_METRICS,
n_pc=5) for x in electrodes
]
if multi_process:
if n_cores is None or n_cores > multiprocessing.cpu_count():
n_cores = multiprocessing.cpu_count() - 1
pool = multiprocessing.get_context('spawn').Pool(n_cores)
for x in clust_objs:
pool.apply_async(x.run, callback=update_pbar)
pool.close()
pool.join()
else:
for x in clust_objs:
res = x.run()
update_pbar(res)
pbar.close()
for rd in rec_dirs:
dat = load_dataset(rd)
dat.process_status['spike_clustering'] = True
dat.process_status['cleanup_clustering'] = False
# dat.save()
# self.save()
print('Clustering Complete\n------------------')
def palatability_identity_calculations(rec_dir, pal_ranks=None,
params=None, shell=False):
warnings.filterwarnings('ignore', category=UserWarning)
warnings.filterwarnings('ignore', category=RuntimeWarning)
dat = load_dataset(rec_dir)
dim = dat.dig_in_mapping
if 'palatability_rank' in dim.columns:
pass
elif pal_ranks is None:
dim = get_palatability_ranks(dim, shell=shell)
else:
dim['palatability_rank'] = dim['name'].map(pal_ranks)
dim = dim.dropna(subset=['palatability_rank'])
dim = dim[dim['palatability_rank'] > 0]
dim = dim.reset_index(drop=True)
num_tastes = len(dim)
taste_names = dim.name.to_list()
trial_list = dat.dig_in_trials.copy()
trial_list = trial_list[[True if x in taste_names else False
for x in trial_list.name]]
num_trials = trial_list.groupby('channel').count()['name'].unique()
if len(num_trials) > 1:
raise ValueError('Unequal number of trials for tastes to used')
else:
num_trials = num_trials[0]
dim['num_trials'] = num_trials
# Get which units to use
unit_table = h5io.get_unit_table(rec_dir)
unit_types = ['Single', 'Multi', 'All', 'Custom']
unit_type = params.get('unit_type')
if unit_type is None:
q = userIO.ask_user('Which units do you want to use for taste '
'discrimination and palatability analysis?',
choices=unit_types,
shell=shell)
unit_type = unit_types[q]
if unit_type == 'Single':
chosen_units = unit_table.loc[unit_table['single_unit'],
'unit_num'].to_list()
elif unit_type == 'Multi':
chosen_units = unit_table.loc[unit_table['single_unit'] == False,
'unit_num'].to_list()
elif unit_type == 'All':
chosen_units = unit_table['unit_num'].to_list()
else:
selection = userIO.select_from_list('Select units to use:',
unit_table['unit_num'],
'Select Units',
multi_select=True)
chosen_units = list(map(int, selection))
num_units = len(chosen_units)
unit_table = unit_table.loc[chosen_units]
# Enter Parameters
if params is None or params.keys() != default_pal_id_params.keys():
params = default_pal_id_params.copy()
params = userIO.confirm_parameter_dict(params,
('Palatability/Identity '
'Calculation Parameters'
'\nTimes in ms'), shell=shell)
win_size = params['window_size']
win_step = params['window_step']
print('Running palatability/identity calculations with parameters:\n%s' %
pt.print_dict(params))
with tables.open_file(dat.h5_file, 'r+') as hf5:
trains_dig_in = hf5.list_nodes('/spike_trains')
time = trains_dig_in[0].array_time[:]
bin_times = np.arange(time[0], time[-1] - win_size + win_step,
win_step)
num_bins = len(bin_times)
palatability = np.empty((num_bins, num_units, num_tastes*num_trials),
dtype=int)
identity = np.empty((num_bins, num_units, num_tastes*num_trials),
dtype=int)
unscaled_response = np.empty((num_bins, num_units, num_tastes*num_trials),
dtype=np.dtype('float64'))
response = np.empty((num_bins, num_units, num_tastes*num_trials),
dtype=np.dtype('float64'))
laser = np.empty((num_bins, num_units, num_tastes*num_trials, 2),
dtype=float)
# Fill arrays with data
print('Filling data arrays...')
onesies = np.ones((num_bins, num_units, num_trials))
for i, row in dim.iterrows():
idx = range(num_trials*i, num_trials*(i+1))
palatability[:, :, idx] = row.palatability_rank * onesies
identity[:, :, idx] = row.channel * onesies
for j, u in enumerate(chosen_units):
for k,t in enumerate(bin_times):
t_idx = np.where((time >= t) & (time <= t+win_size))[0]
unscaled_response[k, j, idx] = \
np.mean(trains_dig_in[i].spike_array[:, u, t_idx],
axis=1)
try:
lasers[k, j, idx] = \
np.vstack((trains_dig_in[i].laser_durations[:],
trains_dig_in[i].laser_onset_lag[:]))
except:
laser[k, j, idx] = np.zeros((num_trials, 2))
# Scaling was not done, so:
response = unscaled_response.copy()
# Make ancillary_analysis node and put in arrays
if '/ancillary_analysis' in hf5:
hf5.remove_node('/ancillary_analysis', recursive=True)
hf5.create_group('/', 'ancillary_analysis')
hf5.create_array('/ancillary_analysis', 'palatability', palatability)
hf5.create_array('/ancillary_analysis', 'identity', identity)
hf5.create_array('/ancillary_analysis', 'laser', laser)
hf5.create_array('/ancillary_analysis', 'scaled_neural_response',
response)
hf5.create_array('/ancillary_analysis', 'window_params',
np.array([win_size, win_step]))
hf5.create_array('/ancillary_analysis', 'bin_times', bin_times)
hf5.create_array('/ancillary_analysis', 'unscaled_neural_response',
unscaled_response)
# for backwards compatibility
hf5.create_array('/ancillary_analysis', 'params',
np.array([win_size, win_step]))
hf5.create_array('/ancillary_analysis', 'pre_stim', np.array(time[0]))
hf5.flush()
# Get unique laser (duration, lag) combinations
print('Organizing trial data...')
unique_lasers = np.vstack(list({tuple(row) for row in laser[0, 0, :, :]}))
unique_lasers = unique_lasers[unique_lasers[:, 1].argsort(), :]
num_conditions = unique_lasers.shape[0]
trials = []
for row in unique_lasers:
tmp_trials = [j for j in range(num_trials * num_tastes)
if np.array_equal(laser[0, 0, j, :], row)]
trials.append(tmp_trials)
trials_per_condition = [len(x) for x in trials]
if not all(x == trials_per_condition[0] for x in trials_per_condition):
raise ValueError('Different number of trials for each laser condition')
trials_per_condition = int(trials_per_condition[0] / num_tastes) #assumes same number of trials per taste per condition
print('Detected:\n %i tastes\n %i laser conditions\n'
' %i trials per condition per taste' %
(num_tastes, num_conditions, trials_per_condition))
trials = np.array(trials)
# Store laser conditions and indices of trials per condition in trial x
# taste space
hf5.create_array('/ancillary_analysis', 'trials', trials)
hf5.create_array('/ancillary_analysis', 'laser_combination_d_l',
unique_lasers)
hf5.flush()
# Taste Similarity Calculation
neural_response_laser = np.empty((num_conditions, num_bins,
num_tastes, num_units,
trials_per_condition),
dtype=np.dtype('float64'))
taste_cosine_similarity = np.empty((num_conditions, num_bins,
num_tastes, num_tastes),
dtype=np.dtype('float64'))
taste_euclidean_distance = np.empty((num_conditions, num_bins,
num_tastes, num_tastes),
dtype=np.dtype('float64'))
# Re-format neural responses from bin x unit x (trial*taste) to
# laser_condition x bin x taste x unit x trial
print('Reformatting data arrays...')
for i, trial in enumerate(trials):
for j, _ in enumerate(bin_times):
for k, _ in dim.iterrows():
idx = np.where((trial >= num_trials*k) &
(trial < num_trials*(k+1)))[0]
neural_response_laser[i, j, k, :, :] = \
response[j, :, trial[idx]].T
# Compute taste cosine similarity and euclidean distances
print('Computing taste cosine similarity and euclidean distances...')
for i, _ in enumerate(trials):
for j, _ in enumerate(bin_times):
for k, _ in dim.iterrows():
for l, _ in dim.iterrows():
taste_cosine_similarity[i, j, k, l] = \
np.mean(cosine_similarity(
neural_response_laser[i, j, k, :, :].T,
neural_response_laser[i, j, l, :, :].T))
taste_euclidean_distance[i, j, k, l] = \
np.mean(cdist(
neural_response_laser[i, j, k, :, :].T,
neural_response_laser[i, j, l, :, :].T,
metric='euclidean'))
hf5.create_array('/ancillary_analysis', 'taste_cosine_similarity',
taste_cosine_similarity)
hf5.create_array('/ancillary_analysis', 'taste_euclidean_distance',
taste_euclidean_distance)
hf5.flush()
# Taste Responsiveness calculations
bin_params = [params['num_comparison_bins'],
params['comparison_bin_size']]
discrim_p = params['discrim_p']
responsive_neurons = []
discriminating_neurons = []
taste_responsiveness = np.zeros((bin_params[0], num_units, 2))
new_bin_times = np.arange(0, np.prod(bin_params), bin_params[1])
baseline = np.where(bin_times < 0)[0]
print('Computing taste responsiveness and taste discrimination...')
for i, t in enumerate(new_bin_times):
places = np.where((bin_times >= t) &
(bin_times <= t+bin_params[1]))[0]
for j, u in enumerate(chosen_units):
# Check taste responsiveness
f, p = f_oneway(np.mean(response[places, j, :], axis=0),
np.mean(response[baseline, j, :], axis=0))
if np.isnan(f):
f = 0.0
p = 1.0
if p <= discrim_p and u not in responsive_neurons:
responsive_neurons.append(u)
taste_responsiveness[i, j, 0] = 1
# Check taste discrimination
taste_idx = [np.arange(num_trials*k, num_trials*(k+1))
for k in range(num_tastes)]
taste_responses = [np.mean(response[places, j, :][:, k], axis=0)
for k in taste_idx]
f, p = f_oneway(*taste_responses)
if np.isnan(f):
f = 0.0
p = 1.0
if p <= discrim_p and u not in discriminating_neurons:
discriminating_neurons.append(u)
responsive_neurons = np.sort(responsive_neurons)
discriminating_neurons = np.sort(discriminating_neurons)
# Write taste responsive and taste discriminating units to text file
save_file = os.path.join(rec_dir, 'discriminative_responsive_neurons.txt')
with open(save_file, 'w') as f:
print('Taste discriminative neurons', file=f)
for u in discriminating_neurons:
print(u, file=f)
print('Taste responsive neurons', file=f)
for u in responsive_neurons:
print(u, file=f)
hf5.create_array('/ancillary_analysis', 'taste_disciminating_neurons',
discriminating_neurons)
hf5.create_array('/ancillary_analysis', 'taste_responsive_neurons',
responsive_neurons)
hf5.create_array('/ancillary_analysis', 'taste_responsiveness',
taste_responsiveness)
hf5.flush()
# Get time course of taste discrimibility
print('Getting taste discrimination time course...')
p_discrim = np.empty((num_conditions, num_bins, num_tastes, num_tastes,
num_units), dtype=np.dtype('float64'))
for i in range(num_conditions):
for j, t in enumerate(bin_times):
for k in range(num_tastes):
for l in range(num_tastes):
for m in range(num_units):
_, p = ttest_ind(neural_response_laser[i, j, k, m, :],
neural_response_laser[i, j, l, m, :],
equal_var = False)
if np.isnan(p):
p = 1.0
p_discrim[i, j, k, l, m] = p
hf5.create_array('/ancillary_analysis', 'p_discriminability',
p_discrim)
hf5.flush()
# Palatability Rank Order calculation (if > 2 tastes)
t_start = params['pal_deduce_start_time']
t_end = params['pal_deduce_end_time']
if num_tastes > 2:
print('Deducing palatability rank order...')
palatability_rank_order_deduction(rec_dir, neural_response_laser,
unique_lasers,
bin_times, [t_start, t_end])
# Palatability calculation
r_spearman = np.zeros((num_conditions, num_bins, num_units))
p_spearman = np.ones((num_conditions, num_bins, num_units))
r_pearson = np.zeros((num_conditions, num_bins, num_units))
p_pearson = np.ones((num_conditions, num_bins, num_units))
f_identity = np.ones((num_conditions, num_bins, num_units))
p_identity = np.ones((num_conditions, num_bins, num_units))
lda_palatability = np.zeros((num_conditions, num_bins))
lda_identity = np.zeros((num_conditions, num_bins))
r_isotonic = np.zeros((num_conditions, num_bins, num_units))
id_pal_regress = np.zeros((num_conditions, num_bins, num_units, 2))
pairwise_identity = np.zeros((num_conditions, num_bins, num_tastes, num_tastes))
print('Computing palatability metrics...')
for i, t in enumerate(trials):
for j in range(num_bins):
for k in range(num_units):
ranks = rankdata(response[j, k, t])
r_spearman[i, j, k], p_spearman[i, j, k] = \
spearmanr(ranks, palatability[j, k, t])
r_pearson[i, j, k], p_pearson[i, j, k] = \
pearsonr(response[j, k, t], palatability[j, k, t])
if np.isnan(r_spearman[i, j, k]):
r_spearman[i, j, k] = 0.0
p_spearman[i, j, k] = 1.0
if np.isnan(r_pearson[i, j, k]):
r_pearson[i, j, k] = 0.0
p_pearson[i, j, k] = 1.0
# Isotonic regression of firing against palatability
model = IsotonicRegression(increasing = 'auto')
model.fit(palatability[j, k, t], response[j, k, t])
r_isotonic[i, j, k] = model.score(palatability[j, k, t],
response[j, k, t])
# Multiple Regression of firing rate against palatability and identity
# Regress palatability on identity
tmp_id = identity[j, k, t].reshape(-1, 1)
tmp_pal = palatability[j, k, t].reshape(-1, 1)
tmp_resp = response[j, k, t].reshape(-1, 1)
model_pi = LinearRegression()
model_pi.fit(tmp_id, tmp_pal)
pi_residuals = tmp_pal - model_pi.predict(tmp_id)
# Regress identity on palatability
model_ip = LinearRegression()
model_ip.fit(tmp_pal, tmp_id)
ip_residuals = tmp_id - model_ip.predict(tmp_pal)
# Regress firing on identity
model_fi = LinearRegression()
model_fi.fit(tmp_id, tmp_resp)
fi_residuals = tmp_resp - model_fi.predict(tmp_id)
# Regress firing on palatability
model_fp = LinearRegression()
model_fp.fit(tmp_pal, tmp_resp)
fp_residuals = tmp_resp - model_fp.predict(tmp_pal)
# Get partial correlation coefficient of response with identity
idp_reg0, p = pearsonr(fp_residuals, ip_residuals)
if np.isnan(idp_reg0):
idp_reg0 = 0.0
idp_reg1, p = pearsonr(fi_residuals, pi_residuals)
if np.isnan(idp_reg1):
idp_reg1 = 0.0
id_pal_regress[i, j, k, 0] = idp_reg0
id_pal_regress[i, j, k, 1] = idp_reg1
# Identity Calculation
samples = []
for _, row in dim.iterrows():
taste = row.channel
samples.append([trial for trial in t
if identity[j, k, trial] == taste])
tmp_resp = [response[j, k, sample] for sample in samples]
f_identity[i, j, k], p_identity[i, j, k] = f_oneway(*tmp_resp)
if np.isnan(f_identity[i, j, k]):
f_identity[i, j, k] = 0.0
p_identity[i, j, k] = 1.0
# Linear Discriminant analysis for palatability
X = response[j, :, t]
Y = palatability[j, 0, t]
test_results = []
c_validator = LeavePOut(1)
for train, test in c_validator.split(X, Y):
model = LDA()
model.fit(X[train, :], Y[train])
tmp = np.mean(model.predict(X[test]) == Y[test])
test_results.append(tmp)
lda_palatability[i, j] = np.mean(test_results)
# Linear Discriminant analysis for identity
Y = identity[j, 0, t]
test_results = []
c_validator = LeavePOut(1)
for train, test in c_validator.split(X, Y):
model = LDA()
model.fit(X[train, :], Y[train])
tmp = np.mean(model.predict(X[test]) == Y[test])
test_results.append(tmp)
lda_identity[i, j] = np.mean(test_results)
# Pairwise Identity Calculation
for ti1, r1 in dim.iterrows():
for ti2, r2 in dim.iterrows():
t1 = r1.channel
t2 = r2.channel
tmp_trials = np.where((identity[j, 0, :] == t1) |
(identity[j, 0, :] == t2))[0]
idx = [trial for trial in t if trial in tmp_trials]
X = response[j, :, idx]
Y = identity[j, 0, idx]
test_results = []
c_validator = StratifiedShuffleSplit(n_splits=10,
test_size=0.25,
random_state=0)
for train, test in c_validator.split(X, Y):
model = GaussianNB()
model.fit(X[train, :], Y[train])
tmp_score = model.score(X[test, :], Y[test])
test_results.append(tmp_score)
pairwise_identity[i, j, ti1, ti2] = np.mean(test_results)
hf5.create_array('/ancillary_analysis', 'r_pearson', r_pearson)
hf5.create_array('/ancillary_analysis', 'r_spearman', r_spearman)
hf5.create_array('/ancillary_analysis', 'p_pearson', p_pearson)
hf5.create_array('/ancillary_analysis', 'p_spearman', p_spearman)
hf5.create_array('/ancillary_analysis', 'lda_palatability', lda_palatability)
hf5.create_array('/ancillary_analysis', 'lda_identity', lda_identity)
hf5.create_array('/ancillary_analysis', 'r_isotonic', r_isotonic)
hf5.create_array('/ancillary_analysis', 'id_pal_regress', id_pal_regress)
hf5.create_array('/ancillary_analysis', 'f_identity', f_identity)
hf5.create_array('/ancillary_analysis', 'p_identity', p_identity)
hf5.create_array('/ancillary_analysis', 'pairwise_NB_identity', pairwise_identity)
hf5.flush()
warnings.filterwarnings('default', category=UserWarning)
warnings.filterwarnings('default', category=RuntimeWarning)
def blech_clust_run(self, data_quality=None, n_cores=None):
'''Write clustering parameters to file and
Run blech_process on each electrode using GNU parallel
Parameters
----------
data_quality : {'clean', 'noisy', None (default)}
set if you want to change the data quality parameters for cutoff
and spike detection before running clustering. These parameters are
automatically set as "clean" during initial parameter setup
accept_params : bool, False (default)
set to True in order to skip popup confirmation of parameters when
running
'''
if data_quality:
tmp = dio.params.load_params('clustering_params',
self.root_dir,
default_keyword=data_quality)
if tmp:
self.clustering_params = tmp
else:
raise ValueError('%s is not a valid data_quality preset. Must '
'be "clean" or "noisy" or None.')
print('\nRunning Blech Clust\n-------------------')
print('Parameters\n%s' % pt.print_dict(self.clustering_params))
# Create folders for saving things within recording dir
data_dir = self.root_dir
directories = [
'spike_waveforms', 'spike_times', 'clustering_results', 'Plots',
'memory_monitor_clustering'
]
for d in directories:
tmp_dir = os.path.join(data_dir, d)
if os.path.exists(tmp_dir):
shutil.rmtree(tmp_dir)
os.mkdir(tmp_dir)
# Set file for clusting log
self.clustering_log = os.path.join(data_dir, 'results.log')
if os.path.exists(self.clustering_log):
os.remove(self.clustering_log)
process_path = os.path.realpath(__file__)
process_path = os.path.join(os.path.dirname(process_path),
'blech_process.py')
em = self.electrode_mapping
if 'dead' in em.columns:
electrodes = em.Electrode[em['dead'] == False].tolist()
else:
electrodes = em.Electrode.tolist()
pbar = tqdm(total=len(electrodes))
results = [(None, None, None)] * (max(electrodes) + 1)
clust_errors = [(x, None) for x in electrodes]
def update_pbar(ans):
if isinstance(ans, tuple) and ans[0] is not None:
results[ans[0]] = ans
else:
print('Unexpected error when clustering an electrode')
pbar.update()
if n_cores is None or n_cores > multiprocessing.cpu_count():
n_cores = multiprocessing.cpu_count() - 1
pool = multiprocessing.Pool(n_cores)
for x in electrodes:
pool.apply_async(blech_clust_process,
args=(x, data_dir, self.clustering_params),
callback=update_pbar)
pool.close()
pool.join()
pbar.close()
print('Electrode Result Cutoff (s)')
cutoffs = {}
clust_res = {}
clustered = []
for x, y, z in results:
if x is None:
continue
clustered.append(x)
print(' {:<13}{:<10}{}'.format(x, y, z))
cutoffs[x] = z
clust_res[x] = y
print('1 - Sucess\n0 - No data or no spikes\n-1 - Error')
em = self.electrode_mapping.copy()
em['cutoff_time'] = em['Electrode'].map(cutoffs)
em['clustering_result'] = em['Electrode'].map(clust_res)
self.electrode_mapping = em.copy()
self.process_status['blech_clust_run'] = True
self.process_status['cleanup_clustering'] = False
dio.h5io.write_electrode_map_to_h5(self.h5_file, em)
self.save()
print('Clustering Complete\n------------------')
return results
def __str__(self):
'''Put all information about dataset in string format
Returns
-------
str : representation of dataset object
'''
out1 = super().__str__()
out = [out1]
out.append('\nObject creation date: ' +
self.dataset_creation_date.strftime('%m/%d/%y'))
if hasattr(self, 'raw_h5_file'):
out.append('Deleted Raw h5 file: ' + self.raw_h5_file)
out.append('h5 File: ' + self.h5_file)
out.append('')
out.append('--------------------')
out.append('Processing Status')
out.append('--------------------')
out.append(pt.print_dict(self.process_status))
out.append('')
if not hasattr(self, 'rec_info'):
return '\n'.join(out)
info = self.rec_info
out.append('--------------------')
out.append('Recording Info')
out.append('--------------------')
out.append(pt.print_dict(self.rec_info))
out.append('')
out.append('--------------------')
out.append('Electrodes')
out.append('--------------------')
out.append(pt.print_dataframe(self.electrode_mapping))
out.append('')
if hasattr(self, 'CAR_electrodes'):
out.append('--------------------')
out.append('CAR Groups')
out.append('--------------------')
headers = ['Group %i' % x for x in range(len(self.CAR_electrodes))]
out.append(pt.print_list_table(self.CAR_electrodes, headers))
out.append('')
if not self.emg_mapping.empty:
out.append('--------------------')
out.append('EMG')
out.append('--------------------')
out.append(pt.print_dataframe(self.emg_mapping))
out.append('')
if info.get('dig_in'):
out.append('--------------------')
out.append('Digital Input')
out.append('--------------------')
out.append(pt.print_dataframe(self.dig_in_mapping))
out.append('')
if info.get('dig_out'):
out.append('--------------------')
out.append('Digital Output')
out.append('--------------------')
out.append(pt.print_dataframe(self.dig_out_mapping))
out.append('')
out.append('--------------------')
out.append('Clustering Parameters')
out.append('--------------------')
out.append(pt.print_dict(self.clustering_params))
out.append('')
out.append('--------------------')
out.append('Spike Array Parameters')
out.append('--------------------')
out.append(pt.print_dict(self.spike_array_params))
out.append('')
out.append('--------------------')
out.append('PSTH Parameters')
out.append('--------------------')
out.append(pt.print_dict(self.psth_params))
out.append('')
out.append('--------------------')
out.append('Palatability/Identity Parameters')
out.append('--------------------')
out.append(pt.print_dict(self.pal_id_params))
out.append('')
return '\n'.join(out)
def validate_data_integrity(rec_dir, verbose=False):
print('Raw Data Validation\n' + '-' * 19)
test_names = [
'file_type', 'recording_info', 'files', 'dropped_packets',
'data_length'
]
number_names = [
'sample_rate', 'dropped_packets', 'missing_files', 'recording_length'
]
tests = dict.fromkeys(test_names, 'NOT TESTED')
numbers = dict.fromkeys(number_names, -1)
file_type = dio.rawIO.get_recording_filetype(rec_dir)
if file_type is None:
file_type_check = 'UNSUPPORTED'
else:
tests['file_type'] = 'PASS'
# Check info.rhd integrity
info_file = os.path.join(rec_dir, 'info.rhd')
try:
rec_info = dio.rawIO.read_rec_info(rec_dir, shell=True)
with open(info_file, 'rb') as f:
info = dio.load_intan_rhd_format.read_header(f)
tests['recording_info'] = 'PASS'
except FileNotFoundError:
test['recording_info'] = 'MISSING'
except Exception as e:
info_size = os.path.getsize(os.path.join(rec_dir, 'info.rhd'))
if info_size == 0:
tests['recording_info'] = 'EMPTY'
else:
tests['recording_info'] = 'FAIL'
print(pt.print_dict(tests, tabs=1))
return tests, numbers
counts = {x: info(x) for x in info.keys() if 'num' in x}
numbers.update(counts)
fs = info['sample_rate']
# Check all files needed are present
files_expected = ['time.dat']
if file_type == 'one file per signal type':
files_expected.append('amplifier.dat')
if rec_info.get('dig_in') is not None:
files_expected.append('digitalin.dat')
if rec_info.get('dig_out') is not None:
files_expected.append('digitalout.dat')
if info['num_auxilary_input_channels'] > 0:
files_expected.append('auxiliary.dat')
elif file_type == 'one file per channel':
for x in info['amplifier_channels']:
files_expected.append('amp-' + x['native_channel_name'] + '.dat')
for x in info['board_dig_in_channels']:
files_expected.append('board-%s.dat' % x['native_channel_name'])
for x in info['board_dig_out_channels']:
files_expected.append('board-%s.dat' % x['native_channel_name'])
for x in info['aux_input_channels']:
files_expected.append('aux-%s.dat' % x['native_channel_name'])
missing_files = []
file_list = os.listdir(rec_dir)
for x in file_expected:
if x not in file_list:
missing_file.append(x)
if len(missing_files) == 0:
tests['files'] = 'PASS'
else:
tests['files'] = 'MISSING'
numbers['missing_files'] = missing_files
# Check time data for dropped packets
time = dio.rawIO.read_time_dat(rec_dir,
sampling_rate=1) # get raw timestamps
numbers['n_samples'] = len(time)
numbers['recording_length'] = float(time[-1]) / fs
expected_time = np.arange(time[0], time[-1] + 1, 1)
missing_timestamps = np.setdiff1d(expected_time, time)
missing_times = np.array([float(x) / fs for x in missing_timestamps])
if len(missing_timestamps) == 0:
tests['dropped_packets'] = 'PASS'
else:
tests['dropped_packets'] = '%i' % len(missing_timestamps)
numbers['dropped_packets'] = missing_times
# Check recording length of each trace
tests['data_traces'] = 'FAIL'
if file_type == 'one file per signal type':
try:
data = dio.rawIO.read_amplifier_dat(rec_dir)
if data is None:
tests['data_traces'] = 'UNREADABLE'
elif data.shape[0] == numbers['n_samples']:
tests['data_traces'] = 'PASS'
else:
tests['data_traces'] = 'CUTOFF'
numbers['data_trace_length (s)'] = data.shape[0] / fs
except:
tests['data_traces'] = 'UNREADABLE'
elif file_type == 'one file per channel':
chan_info = pd.DataFrame(columns=['port', 'channel', 'n_samples'])
lengths = []
min_samples = numbers['n_samples']
max_samples = number['n_samples']
for x in info['amplifier_channels']:
fn = os.path.join(rec_dir, 'amp-%s.dat' % x['native_channel_name'])
if os.path.basename(fn) in missing_files:
continue
data = dio.rawIO.read_one_channel_file(fn)
lengths.append((x['native_channel_name'], data.shape[0]))
if data.shape[0] < min_samples:
min_samples = data.shape[0]
if data.shape[0] > max_samples:
max_samples = data.shape[0]
if min_samples == max_samples:
tests['data_traces'] = 'PASS'
else:
test['data_traces'] = 'CUTOFF'
numbers['max_recording_length (s)'] = max_samples / fs
numbers['min_recording_length (s)'] = min_samples / fs
