def _bids2nipypeinfo(in_file,
events_file,
regressors_file,
regressors_names=None,
motion_columns=None,
decimals=3,
amplitude=1.0):
from pathlib import Path
import numpy as np
import pandas as pd
from nipype.interfaces.base.support import Bunch
removeTR = 4
# Process the events file
events = pd.read_csv(events_file, sep=r'\s+')
bunch_fields = ['onsets', 'durations', 'amplitudes']
if not motion_columns:
from itertools import product
motion_columns = [
'_'.join(v) for v in product(('trans', 'rot'), 'xyz')
]
out_motion = Path('motion.par').resolve()
regress_data = pd.read_csv(regressors_file, sep=r'\s+')
np.savetxt(out_motion, regress_data[motion_columns].values, '%g')
if regressors_names is None:
regressors_names = sorted(
set(regress_data.columns) - set(motion_columns))
if regressors_names:
bunch_fields += ['regressor_names']
bunch_fields += ['regressors']
runinfo = Bunch(scans=in_file,
conditions=list(set(events.trial_type_N.values)),
**{k: []
for k in bunch_fields})
for condition in runinfo.conditions:
event = events[events.trial_type_N.str.match(condition)]
runinfo.onsets.append(
np.round(event.onset.values - removeTR, 3).tolist()
) # added -removeTR to align to the onsets after removing X number of TRs from the scan
runinfo.durations.append(np.round(event.duration.values, 3).tolist())
if 'amplitudes' in events.columns:
runinfo.amplitudes.append(
np.round(event.amplitudes.values, 3).tolist())
else:
runinfo.amplitudes.append([amplitude] * len(event))
if 'regressor_names' in bunch_fields:
runinfo.regressor_names = regressors_names
runinfo.regressors = regress_data[regressors_names].fillna(0.0).values[
removeTR:, ].T.tolist() # adding removeTR to cut the first rows
return [runinfo], str(out_motion)
def _bids2nipypeinfo(in_file, events_file, regressors_file,
regressors_names=None,
motion_columns=None,
decimals=3, amplitude=1.0, del_scan=10):
from pathlib import Path
import numpy as np
import pandas as pd
from nipype.interfaces.base.support import Bunch
# Process the events file
events = pd.read_csv(events_file, sep=r'\s+')
bunch_fields = ['onsets', 'durations', 'amplitudes']
if not motion_columns:
from itertools import product
motion_columns = ['_'.join(v) for v in product(('trans', 'rot'), 'xyz')]
out_motion = Path('motion.par').resolve()
regress_data = pd.read_csv(regressors_file, sep=r'\s+')
np.savetxt(out_motion, regress_data[motion_columns].fillna(0.0).values[del_scan:,], '%g')
# np.savetxt(out_motion, regress_data[motion_columns].fillna(0.0).values, '%g')
if regressors_names is None:
regressors_names = sorted(set(regress_data.columns) - set(motion_columns))
if regressors_names:
bunch_fields += ['regressor_names']
bunch_fields += ['regressors']
domain = list(set(events.condition.values))[0] # domain of this task run, should be only one, 'Mon' or 'Med'
trial_types = list(set(events.trial_type.values))
# add parametric modulator
bunch_fields += ['pmod']
runinfo = Bunch(
scans=in_file,
conditions=[domain + '_' + trial_type for trial_type in trial_types],
# conditions = ['Med_amb', 'Med_risk', 'Mon_amb', 'Mon_risk'],
**{k: [] for k in bunch_fields})
for condition in runinfo.conditions:
event = events[events.trial_type.str.match(condition[4:])]
runinfo.onsets.append(np.round(event.onset.values - del_scan + 1, 3).tolist()) # take out the first several deleted scans
runinfo.durations.append(np.round(event.duration.values, 3).tolist())
# parametric modulator
runinfo.pmod.append(Bunch(
name = [condition + '_reward_prob'], #name of modulator for each condition
param = [np.round(event.reward_prob.values, 3).tolist()], # values of modulator for each condition
poly = [1] #degree of modulation, 1-linear
))
if 'amplitudes' in events.columns:
runinfo.amplitudes.append(np.round(event.amplitudes.values, 3).tolist())
else:
runinfo.amplitudes.append([amplitude] * len(event))
# response predictor regardless of condition
runinfo.conditions.append('Resp')
# response predictor when there is a button press
resp_mask = events.resp != 2
resp_onset= np.round(events.resp_onset.values[resp_mask] - del_scan + 1, 3).tolist()
runinfo.onsets.append(resp_onset)
runinfo.durations.append([0] * len(resp_onset))
runinfo.amplitudes.append([amplitude] * len(resp_onset))
# no parametric modulator for response
runinfo.pmod.append(None)
if 'regressor_names' in bunch_fields:
runinfo.regressor_names = regressors_names
runinfo.regressors = regress_data[regressors_names].fillna(0.0).values[del_scan:,].T.tolist()
return runinfo, str(out_motion)
from itertools import product
motion_columns = ['_'.join(v) for v in product(('trans', 'rot'), 'xyz')]
out_motion = Path('motion.par').resolve()
regress_data = pd.read_csv(regressors_file, sep=r'\s+')
np.savetxt(out_motion, regress_data[motion_columns].values, '%g')
if regressors_names is None:
regressors_names = sorted(set(regress_data.columns) - set(motion_columns))
if regressors_names:
bunch_fields += ['regressor_names']
bunch_fields += ['regressors']
runinfo = Bunch(
scans=in_file,
conditions=list(set(events.trial_type.values)),
**{k: [] for k in bunch_fields})
for condition in runinfo.conditions:
event = events[events.trial_type.str.match(condition)]
runinfo.onsets.append(np.round(event.onset.values, 3).tolist())
runinfo.durations.append(np.round(event.duration.values, 3).tolist())
if 'amplitudes' in events.columns:
runinfo.amplitudes.append(np.round(event.amplitudes.values, 3).tolist())
else:
runinfo.amplitudes.append([amplitude] * len(event))
if 'regressor_names' in bunch_fields:
runinfo.regressor_names = regressors_names
runinfo.regressors = regress_data[regressors_names].fillna(0.0).values.T.tolist()
def _bids2nipypeinfo(in_file,
events_file,
regressors_file,
regressors_names=None,
motion_columns=None,
decimals=3,
amplitude=1.0,
del_scan=10):
from pathlib import Path
import numpy as np
import pandas as pd
from nipype.interfaces.base.support import Bunch
# Process the events file
events = pd.read_csv(events_file, sep=r'\s+')
bunch_fields = ['onsets', 'durations', 'amplitudes']
if not motion_columns:
from itertools import product
motion_columns = [
'_'.join(v) for v in product(('trans', 'rot'), 'xyz')
]
out_motion = Path('motion.par').resolve()
regress_data = pd.read_csv(regressors_file, sep=r'\s+')
np.savetxt(out_motion,
regress_data[motion_columns].fillna(0.0).values[del_scan:, ],
'%g')
# np.savetxt(out_motion, regress_data[motion_columns].fillna(0.0).values, '%g')
if regressors_names is None:
regressors_names = sorted(
set(regress_data.columns) - set(motion_columns))
if regressors_names:
bunch_fields += ['regressor_names']
bunch_fields += ['regressors']
domain = list(set(events.condition.values))[
0] # domain of this task run, should be only one, 'Mon' or 'Med'
trial_types = list(set(events.trial_type.values))
# outcome_levels = list(set(events.vals.values))
outcome_levels = {'0': 5, '1': 8, '2': 12, '3': 25}
conds = [domain + '_' + trial_type
for trial_type in trial_types] # e.g. ['Med_risk', 'Med_ambig']
conditions = []
for cond in conds:
for outcome_level in list(outcome_levels.keys()):
conditions.append(cond + '_' + outcome_level)
runinfo = Bunch(
scans=in_file,
conditions=conditions, # should be 8 conditions (2 x 4)
# conditions = ['Med_amb', 'Med_risk', 'Mon_amb', 'Mon_risk'],
**{k: []
for k in bunch_fields})
for condition in runinfo.conditions:
event = events[
(events.trial_type.str.match(condition[4:len(condition) - 2]))
& (events.vals == outcome_levels[condition[-1:]])]
runinfo.onsets.append(
np.round(event.onset.values - del_scan + 1,
3).tolist()) # take out the first several deleted scans
runinfo.durations.append(np.round(event.duration.values, 3).tolist())
if 'amplitudes' in events.columns:
runinfo.amplitudes.append(
np.round(event.amplitudes.values, 3).tolist())
else:
runinfo.amplitudes.append([amplitude] * len(event))
# if domain == condition[:3]:
# event = events[events.trial_type.str.match(condition[4:])]
# runinfo.onsets.append(np.round(event.onset.values - del_scan + 1, 3).tolist()) # take out the first several deleted scans
# runinfo.durations.append(np.round(event.duration.values, 3).tolist())
# if 'amplitudes' in events.columns:
# runinfo.amplitudes.append(np.round(event.amplitudes.values, 3).tolist())
# else:
# runinfo.amplitudes.append([amplitude] * len(event))
# else: # empty conditions
# runinfo.onsets.append([])
# runinfo.durations.append([])
# runinfo.amplitudes.append([])
# delete empty condition, if any
cond_idx = 0
while cond_idx < len(runinfo.conditions):
if not runinfo.onsets[cond_idx]:
runinfo.conditions.pop(cond_idx)
runinfo.onsets.pop(cond_idx)
runinfo.durations.pop(cond_idx)
runinfo.amplitudes.pop(cond_idx)
else:
cond_idx += 1
# response predictor regardless of condition
runinfo.conditions.append('Resp')
# response predictor when there is a button press
resp_mask = events.resp != 2
resp_onset = np.round(events.resp_onset.values[resp_mask] - del_scan + 1,
3).tolist()
runinfo.onsets.append(resp_onset)
runinfo.durations.append([0] * len(resp_onset))
runinfo.amplitudes.append([amplitude] * len(resp_onset))
if 'regressor_names' in bunch_fields:
runinfo.regressor_names = regressors_names
runinfo.regressors = regress_data[regressors_names].fillna(0.0).values[
del_scan:, ].T.tolist()
return runinfo, str(out_motion)
from nipype.interfaces.base.support import Bunch
from ..path import findpaths
A = "/tmp/a.txt" # TODO make this more elegant with a tmp_dir
B = "/tmp/b.txt"
@pytest.mark.timeout(60)
@pytest.mark.parametrize("obj", [[A, B], (A, B), {A, B}, {
"a": A,
"b": B
}, {
"x": {
"y": [A, B]
}
},
Bunch(a=A, b=B),
Bunch(x=[A, B])])
def test_findpaths(tmp_path, obj):
os.chdir(str(tmp_path))
for fname in [A, B]:
Path(fname).touch()
assert set(findpaths(obj)) == set([A, B])
for fname in [A, B]:
Path(fname).unlink()
def _bids2nipypeinfo(in_file,
events_file,
regressors_file,
regressors_names=None,
motion_columns=None,
decimals=3,
amplitude=1.0,
del_scan=10):
from pathlib import Path
import numpy as np
import pandas as pd
from nipype.interfaces.base.support import Bunch
# Process the events file
# events_file = '/home/rj299/project/mdm_analysis/output/event_files/sub-2073_task-4_cond_v3.csv'
events = pd.read_csv(events_file, sep=r'\s+')
bunch_fields = ['onsets', 'durations', 'amplitudes']
if not motion_columns:
from itertools import product
motion_columns = [
'_'.join(v) for v in product(('trans', 'rot'), 'xyz')
]
out_motion = Path('motion.par').resolve()
regress_data = pd.read_csv(regressors_file, sep=r'\s+')
np.savetxt(out_motion,
regress_data[motion_columns].fillna(0.0).values[del_scan:, ],
'%g')
# np.savetxt(out_motion, regress_data[motion_columns].fillna(0.0).values, '%g')
if regressors_names is None:
regressors_names = sorted(
set(regress_data.columns) - set(motion_columns))
if regressors_names:
bunch_fields += ['regressor_names']
bunch_fields += ['regressors']
domain = list(set(events.condition.values))[
0] # domain of this task run, should be only one, 'Mon' or 'Med'
trial_types = list(set(events.trial_type.values))
trial_types.sort() # separate trials by reward magnitude
# read unique risk levels
risk_levels_percent = events.probs.values * 100
risk_levels = list(set(risk_levels_percent.astype(int)))
risk_levels.sort()
# read unique ambiguity levels
amb_levels_percent = events.ambigs.values * 100
amb_levels = list(set(amb_levels_percent.astype(int)))
amb_levels.sort()
amb_levels.pop(
0) # get rid of the ambgiuity 0 level (because those are risky trials)
runinfo = Bunch(
scans=in_file,
conditions=[
domain + '_' + str(trial_types[0]) + '_' + str(amb_level)
for amb_level in amb_levels
] + [
domain + '_' + str(trial_types[1]) + '_' + str(risk_level)
for risk_level in risk_levels
],
# conditions = ['Med_amb_24', 'Med_amb_50', 'Med_amb_74',
# 'Med_risk_25', 'Med_risk_50','Med_risk_75',
# 'Mon_amb_24', 'Mon_amb_50', 'Mon_amb_74',
# 'Mon_risk_25', 'Mon_risk_50','Mon_risk_75']
**{k: []
for k in bunch_fields})
for condition in runinfo.conditions:
if condition[4:-3] == 'amb':
event = events[(events.trial_type == condition[4:-3])
& (events.ambigs * 100 == int(condition[-2:]))]
else:
event = events[(events.trial_type == condition[4:-3])
& (events.probs * 100 == int(condition[-2:]))]
runinfo.onsets.append(
np.round(event.onset.values - del_scan + 1,
3).tolist()) # take out the first several deleted scans
runinfo.durations.append(np.round(event.duration.values, 3).tolist())
if 'amplitudes' in events.columns:
runinfo.amplitudes.append(
np.round(event.amplitudes.values, 3).tolist())
else:
runinfo.amplitudes.append([amplitude] * len(event))
# response predictor regardless of condition
runinfo.conditions.append('Resp')
# response predictor when there is a button press
resp_mask = events.resp != 2
resp_onset = np.round(events.resp_onset.values[resp_mask] - del_scan + 1,
3).tolist()
runinfo.onsets.append(resp_onset)
runinfo.durations.append([0] * len(resp_onset))
runinfo.amplitudes.append([amplitude] * len(resp_onset))
# delete empty condition
runinfo.conditions = [
x for (x, y) in zip(runinfo.conditions, runinfo.amplitudes) if y
]
runinfo.amplitudes = [x for x in runinfo.amplitudes if x]
runinfo.durations = [x for x in runinfo.durations if x]
runinfo.onsets = [x for x in runinfo.onsets if x]
# if domain == condition[:3]:
# event = events[events.trial_type.str.match(condition[4:])]
# runinfo.onsets.append(np.round(event.onset.values - del_scan + 1, 3).tolist()) # take out the first several deleted scans
# runinfo.durations.append(np.round(event.duration.values, 3).tolist())
# if 'amplitudes' in events.columns:
# runinfo.amplitudes.append(np.round(event.amplitudes.values, 3).tolist())
# else:
# runinfo.amplitudes.append([amplitude] * len(event))
# else: # empty conditions
# runinfo.onsets.append([])
# runinfo.durations.append([])
# runinfo.amplitudes.append([])
if 'regressor_names' in bunch_fields:
runinfo.regressor_names = regressors_names
runinfo.regressors = regress_data[regressors_names].fillna(0.0).values[
del_scan:, ].T.tolist()
return runinfo, str(out_motion)
