def _run_interface(self, runtime):
from nltools.data import Brain_Data
import os
in_file = self.inputs.in_file
mask = self.inputs.mask
low_pass = self.inputs.low_pass_cutoff
high_pass = self.inputs.high_pass_cutoff
TR = self.inputs.sampling_rate
if low_pass == 0:
low_pass = None
if high_pass == 0:
high_pass = None
dat = Brain_Data(in_file, mask=mask)
# Handle no filtering
if low_pass or high_pass:
dat = dat.filter(sampling_rate=TR,
low_pass=low_pass,
high_pass=high_pass)
# Generate output file name
out_file = os.path.split(in_file)[-1].split(
'.nii.gz')[0] + '_filtered.nii.gz'
dat.write(out_file)
self._out_file = out_file
runtime.returncode = 0
return runtime
def test_roc(tmpdir):
sim = simulator.Simulator()
r = 10
sigma = .1
y = [0, 1]
n_reps = 10
# output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=None)
dat = Brain_Data(data=sim.data,Y=pd.DataFrame(sim.y))
algorithm = 'svm'
# output_dir = str(tmpdir)
# cv = {'type': 'kfolds', 'n_folds': 5, 'subject_id': sim.rep_id}
extra = {'kernel': 'linear'}
output = dat.predict(algorithm='svm', plot=False, **extra)
# Single-Interval
roc = analysis.Roc(input_values=output['yfit_all'], binary_outcome=output['Y'] == 1)
roc.calculate()
roc.summary()
assert roc.accuracy == 1
# Forced Choice
binary_outcome = output['Y'] == 1
forced_choice = range(len(binary_outcome)/2) + range(len(binary_outcome)/2)
forced_choice = forced_choice.sort()
roc_fc = analysis.Roc(input_values=output['yfit_all'], binary_outcome=binary_outcome, forced_choice=forced_choice)
roc_fc.calculate()
assert roc_fc.accuracy == 1
assert roc_fc.accuracy == roc_fc.auc == roc_fc.sensitivity == roc_fc.specificity
def test_roc(tmpdir, sim):
r = 10
sigma = .1
y = [0, 1]
n_reps = 10
output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=None)
dat = Brain_Data(data=sim.data, Y=pd.DataFrame(sim.y))
algorithm = 'svm'
output_dir = str(tmpdir)
# cv = {'type': 'kfolds', 'n_folds': 5, 'subject_id': sim.rep_id}
extra = {'kernel': 'linear'}
output = dat.predict(algorithm='svm', **extra)
# predict = analysis.Predict(sim.data, sim.y, algorithm=algorithm,
# output_dir=output_dir,
# cv_dict=cv,
# **extra)
# predict.predict() # Save_Plot isn't working for SVM analysis, planning on deprecating analysis.Predict at some point, so not a big deal
# predict.predict(save_plot=False)
# Single-Interval
roc = analysis.Roc(input_values=output['yfit_all'],
binary_outcome=output['Y'] == 1)
# roc = analysis.Roc(
# input_values=predict.yfit_xval, binary_outcome=np.array(sim.y) == 1)
roc.plot()
roc.summary()
assert roc.accuracy == 1
def expand_mask(mask, custom_mask=None):
"""expand a mask with multiple integers into separate binary masks
Args:
mask: nibabel or Brain_Data instance
custom_mask: nibabel instance or string to file path; optional
Returns:
out: Brain_Data instance of multiple binary masks
"""
from nltools.data import Brain_Data
if isinstance(mask, nib.Nifti1Image):
mask = Brain_Data(mask, mask=custom_mask)
if not isinstance(mask, Brain_Data):
raise ValueError("Make sure mask is a nibabel or Brain_Data instance.")
mask.data = np.round(mask.data).astype(int)
tmp = []
for i in np.nonzero(np.unique(mask.data))[0]:
tmp.append((mask.data == i) * 1)
out = mask.empty()
out.data = np.array(tmp)
return out
def test_roc(tmpdir):
sim = simulator.Simulator()
r = 10
sigma = .1
y = [0, 1]
n_reps = 10
# output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=None)
dat = Brain_Data(data=sim.data, Y=pd.DataFrame(sim.y))
algorithm = 'svm'
# output_dir = str(tmpdir)
# cv = {'type': 'kfolds', 'n_folds': 5, 'subject_id': sim.rep_id}
extra = {'kernel': 'linear'}
output = dat.predict(algorithm='svm', plot=False, **extra)
# Single-Interval
roc = analysis.Roc(input_values=output['yfit_all'],
binary_outcome=output['Y'] == 1)
# roc = analysis.Roc(
# input_values=predict.yfit_xval, binary_outcome=np.array(sim.y) == 1)
roc.calculate()
# roc.plot()
roc.summary()
assert roc.accuracy == 1
def test_roc(tmpdir, sim):
r = 10
sigma = .1
y = [0, 1]
n_reps = 10
output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=None)
dat = Brain_Data(data=sim.data,Y=pd.DataFrame(sim.y))
algorithm = 'svm'
output_dir = str(tmpdir)
# cv = {'type': 'kfolds', 'n_folds': 5, 'subject_id': sim.rep_id}
extra = {'kernel': 'linear'}
output = dat.predict(algorithm='svm',**extra)
# predict = analysis.Predict(sim.data, sim.y, algorithm=algorithm,
# output_dir=output_dir,
# cv_dict=cv,
# **extra)
# predict.predict() # Save_Plot isn't working for SVM analysis, planning on deprecating analysis.Predict at some point, so not a big deal
# predict.predict(save_plot=False)
# Single-Interval
roc = analysis.Roc(input_values=output['yfit_all'], binary_outcome=output['Y'] == 1)
# roc = analysis.Roc(
# input_values=predict.yfit_xval, binary_outcome=np.array(sim.y) == 1)
roc.plot()
roc.summary()
assert roc.accuracy == 1
def collapse_mask(mask, auto_label=True, custom_mask=None):
"""collapse separate masks into one mask with multiple integers
overlapping areas are ignored
Args:
mask: nibabel or Brain_Data instance
custom_mask: nibabel instance or string to file path; optional
Returns:
out: Brain_Data instance of a mask with different integers indicating
different masks
"""
from nltools.data import Brain_Data
if not isinstance(mask, Brain_Data):
if isinstance(mask, nib.Nifti1Image):
mask = Brain_Data(mask, mask=custom_mask)
else:
raise ValueError("Make sure mask is a nibabel or Brain_Data "
"instance.")
if len(mask.shape()) > 1:
if len(mask) > 1:
out = mask.empty()
# Create list of masks and find any overlaps
m_list = []
for x in range(len(mask)):
m_list.append(mask[x].to_nifti())
intersect = intersect_masks(m_list, threshold=1, connected=False)
intersect = Brain_Data(
nib.Nifti1Image(np.abs(intersect.get_data() - 1),
intersect.get_affine()),
mask=custom_mask,
)
merge = []
if auto_label:
# Combine all masks into sequential order
# ignoring any areas of overlap
for i in range(len(m_list)):
merge.append(
np.multiply(
Brain_Data(m_list[i], mask=custom_mask).data,
intersect.data) * (i + 1))
out.data = np.sum(np.array(merge).T, 1).astype(int)
else:
# Collapse masks using value as label
for i in range(len(m_list)):
merge.append(
np.multiply(
Brain_Data(m_list[i], mask=custom_mask).data,
intersect.data))
out.data = np.sum(np.array(merge).T, 1)
return out
else:
warnings.warn("Doesn't need to be collapased")
def create_data(self, levels, sigma, radius = 5, center = None, reps = 1, output_dir = None):
""" create simulated data with integers
Args:
levels: vector of intensities or class labels
sigma: amount of noise to add
radius: vector of radius. Will create multiple spheres if len(radius) > 1
center: center(s) of sphere(s) of the form [px, py, pz] or [[px1, py1, pz1], ..., [pxn, pyn, pzn]]
reps: number of data repetitions useful for trials or subjects
output_dir: string path of directory to output data. If None, no data will be written
**kwargs: Additional keyword arguments to pass to the prediction algorithm
"""
# Create reps
nlevels = len(levels)
y = levels
rep_id = [1] * len(levels)
for i in range(reps - 1):
y = y + levels
rep_id.extend([i+2] * nlevels)
# Initialize Spheres with options for multiple radii and centers of the spheres (or just an int and a 3D list)
A = self.n_spheres(radius, center)
#for each intensity
A_list = []
for i in y:
A_list.append(np.multiply(A, i))
#generate a different gaussian noise profile for each mask
mu = 0 #values centered around 0
N_list = []
for i in range(len(y)):
N_list.append(self.normal_noise(mu, sigma))
#add noise and signal together, then convert to nifti files
NF_list = []
for i in range(len(y)):
NF_list.append(self.to_nifti(np.add(N_list[i], A_list[i]) ))
NF_list = Brain_Data(NF_list)
# Assign variables to object
self.data = NF_list
self.y = pd.DataFrame(data=y)
self.rep_id = pd.DataFrame(data=rep_id)
dat = self.data
dat.Y = self.y
# Write Data to files if requested
if output_dir is not None and isinstance(output_dir, six.string_types):
NF_list.write(os.path.join(output_dir,'data.nii.gz'))
self.y.to_csv(os.path.join(output_dir, 'y.csv'), index=None,header=False)
self.rep_id.to_csv(os.path.join(output_dir, 'rep_id.csv'), index=None,header=False)
return dat
def test_load(tmpdir):
sim = Simulator()
sigma = 1
y = [0, 1]
n_reps = 3
output_dir = str(tmpdir)
dat = sim.create_data(y, sigma, reps=n_reps, output_dir=output_dir)
# if MNI_Template["resolution"] == '2mm':
# shape_3d = (91, 109, 91)
# shape_2d = (6, 238955)
# elif MNI_Template["resolution"] == '3mm':
# shape_3d = (60, 72, 60)
# shape_2d = (6, 71020)
y = pd.read_csv(os.path.join(str(tmpdir.join('y.csv'))), header=None, index_col=None)
# holdout = pd.read_csv(os.path.join(str(tmpdir.join('rep_id.csv'))), header=None, index_col=None)
# Test load list of 4D images
file_list = [str(tmpdir.join('data.nii.gz')), str(tmpdir.join('data.nii.gz'))]
dat = Brain_Data(file_list)
dat = Brain_Data([nb.load(x) for x in file_list])
# Test load list
dat = Brain_Data(data=str(tmpdir.join('data.nii.gz')), Y=y)
# Test Write
dat.write(os.path.join(str(tmpdir.join('test_write.nii'))))
assert Brain_Data(os.path.join(str(tmpdir.join('test_write.nii'))))
def check_brain_data(data, mask=None):
"""Check if data is a Brain_Data Instance."""
from nltools.data import Brain_Data
if not isinstance(data, Brain_Data):
if isinstance(data, nib.Nifti1Image):
data = Brain_Data(data, mask=mask)
else:
raise ValueError("Make sure data is a Brain_Data instance.")
else:
if mask is not None:
data = data.apply_mask(mask)
return data
def similarity(image_list,
weight_map_filename,
file_path_key='resampled_file'):
tic = time.time() # Start Timer
weight_map = nb.load(weight_map_filename)
file_path_list = [item[file_path_key] for item in image_list]
dat = Brain_Data(data=file_path_list)
r = dat.similarity(weight_map)
log.info("Elapsed: %.2f seconds", (time.time() - tic)) # Stop timer
return {'correlation': set_correlation(r, image_list)}
def similarity(image_list, weight_map_filename,
file_path_key='resampled_file'):
tic = time.time() # Start Timer
weight_map = nb.load(weight_map_filename)
file_path_list = [item[file_path_key] for item in image_list]
dat = Brain_Data(data=file_path_list)
r = dat.similarity(weight_map)
log.info("Elapsed: %.2f seconds", (time.time() - tic)) # Stop timer
return {
'correlation': set_correlation(r, image_list)
}
def test_roi_to_brain():
s1 = create_sphere([15, 10, -8], radius=10)
s2 = create_sphere([-15, 10, -8], radius=10)
s3 = create_sphere([0, -15, -8], radius=10)
masks = Brain_Data([s1, s2, s3])
d = [1, 2, 3]
m = roi_to_brain(d, masks)
assert np.all([np.any(m.data == x) for x in d])
d = pd.Series([1.1, 2.1, 3.1])
m = roi_to_brain(d, masks)
assert np.all([np.any(m.data == x) for x in d])
d = np.array([1, 2, 3])
m = roi_to_brain(d, masks)
assert np.all([np.any(m.data == x) for x in d])
d = pd.DataFrame([np.ones(10) * x for x in [1, 2, 3]])
m = roi_to_brain(d, masks)
assert len(m) == d.shape[1]
assert np.all([np.any(m[0].data == x) for x in d[0]])
d = np.array([np.ones(10) * x for x in [1, 2, 3]])
m = roi_to_brain(d, masks)
assert len(m) == d.shape[1]
assert np.all([np.any(m[0].data == x) for x in d[0]])
def reload(self):
self.layout = BIDSLayout(self.root, derivatives=True)
self.tr = self.layout.get_tr()
with open(self.layout._get_unique(scope=self.name, suffix="pipeline").path) as file:
pipeline = json.load(file)
sys.path.append(os.path.dirname(self.layout._get_unique(
scope=self.name, suffix="pipeline").path))
self.masks = dict()
for mask, mask_path in pipeline["Masks"].items():
if not os.path.isabs(mask_path):
mask_path = join(self.root, "derivatives",
self.name, mask_path)
self.masks[mask] = Brain_Data(mask_path)
# Set up the process dictionary
self.processes = dict()
for process in pipeline["Processes"]:
if not os.path.isabs(process["Source"]):
process["Source"] = join(self.root, "derivatives",
self.name, process["Source"])
head, tail = os.path.split(os.path.abspath(process["Source"]))
if tail.endswith(".py"):
tail = tail[:-3]
else:
raise TypeError(f"{tail} is not a Python script.")
sys.path.append(head)
self.processes[process["Name"]] = Process(
key=process["Readable"], process=getattr(__import__(tail), process["Name"]))
sys.path.remove(head)
def fetch_emotion_ratings(data_dir=None, resume=True, verbose=1):
'''Download and loads emotion rating dataset from neurovault
Args:
data_dir: (string, optional). Path of the data directory. Used to
force data storage in a specified location. Default: None
n_subjects: (int, optional) Number of subjects, from 1 to 6.
NOTE: n_subjects is deprecated from 0.2.6 and will be
removed in 0.3 Use `subjects` instead.
subjects : (list or int, optional) Either a list of subjects or the
number of subjects to load, from 1 to 6. By default,
2nd subject will be loaded. Empty list returns no subject
data
Returns:
'''
collection = 1964
dataset_name = 'chang2015_emotion_ratings'
data_dir = _get_dataset_dir(dataset_name,
data_dir=data_dir,
verbose=verbose)
metadata, files = download_collection(collection=collection,
data_dir=data_dir,
resume=resume,
verbose=verbose)
return Brain_Data(data=files, X=metadata)
def roi_to_brain(data, mask_x):
'''
Args:
data: Pandas series or dataframe of ROI by observation
mask_x: an expanded binary mask
Returns:
Brain_Data instance
'''
def series_to_brain(data, mask_x):
'''Converts a pandas series of ROIs to a Brain_Data instance. Index must correspond to ROI index'''
if not isinstance(data, pd.Series):
raise ValueError('Data must be a pandas series')
if len(mask_x) != len(data):
raise ValueError(
'Data must have the same number of rows as mask has ROIs.')
return Brain_Data([mask_x[x] * data[x] for x in data.keys()]).sum()
if len(mask_x) != data.shape[0]:
raise ValueError(
'Data must have the same number of rows as mask has ROIs.')
if isinstance(data, pd.Series):
return series_to_brain(data, mask_x)
elif isinstance(data, pd.DataFrame):
return Brain_Data(
[series_to_brain(data[x], mask_x) for x in data.keys()])
def roi_to_brain(data, mask_x):
''' This function will create convert an expanded binary mask of ROIs
(see expand_mask) based on a vector of of values. The dataframe of values
must correspond to ROI numbers.
This is useful for populating a parcellation scheme by a vector of Values
Args:
data: Pandas series or dataframe of ROI by observation
mask_x: an expanded binary mask
Returns:
out: (Brain_Data) Brain_Data instance where each ROI is now populated
with a value
'''
from nltools.data import Brain_Data
def series_to_brain(data, mask_x):
'''Converts a pandas series of ROIs to a Brain_Data instance. Index must correspond to ROI index'''
if not isinstance(data, pd.Series):
raise ValueError('Data must be a pandas series')
if len(mask_x) != len(data):
raise ValueError(
'Data must have the same number of rows as mask has ROIs.')
return Brain_Data([mask_x[x] * data[x] for x in data.keys()]).sum()
if len(mask_x) != data.shape[0]:
raise ValueError(
'Data must have the same number of rows as mask has ROIs.')
if isinstance(data, pd.Series):
return series_to_brain(data, mask_x)
elif isinstance(data, pd.DataFrame):
return Brain_Data(
[series_to_brain(data[x], mask_x) for x in data.keys()])
def test_check_brain_data(sim_brain_data):
mask = Brain_Data(create_sphere([15, 10, -8], radius=10))
a = check_brain_data(sim_brain_data)
assert isinstance(a, Brain_Data)
b = check_brain_data(sim_brain_data, mask=mask)
assert isinstance(b, Brain_Data)
assert b.shape()[1] == np.sum(mask.data==1)
def test_groupby(tmpdir):
# Simulate Brain Data
sim = Simulator()
r = 10
sigma = 1
y = [0, 1]
n_reps = 3
output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=output_dir)
s1 = create_sphere([12, 10, -8], radius=r)
s2 = create_sphere([22, -2, -22], radius=r)
mask = Brain_Data([s1, s2])
y = pd.read_csv(os.path.join(str(tmpdir.join('y.csv'))),
header=None,
index_col=None)
data = Brain_Data(glob.glob(str(tmpdir.join('data.nii.gz'))), Y=y)
data.X = pd.DataFrame(
{
'Intercept': np.ones(len(data.Y)),
'X1': np.array(data.Y).flatten()
},
index=None)
dat = Groupby(data, mask)
# Test length
assert len(dat) == len(mask)
# Test Index
assert isinstance(dat[1], Brain_Data)
# Test apply
mn = dat.apply('mean')
assert len(dat) == len(mn)
# assert mn[0].mean() > mn[1].mean() #JC edit: it seems this check relies on chance from simulated data
assert mn[1].shape() == np.sum(mask[1].data == 1)
reg = dat.apply('regress')
assert len(dat) == len(mn)
# r = dict([(x,reg[x]['beta'][1]) for x in reg.iterkeys()])
# Test combine
combine_mn = dat.combine(mn)
assert len(combine_mn.shape()) == 1
def test_groupby_aggregate(sim_brain_data):
s1 = create_sphere([12, 10, -8], radius=10)
s2 = create_sphere([22, -2, -22], radius=10)
mask = Brain_Data([s1, s2])
d = sim_brain_data.groupby(mask)
assert isinstance(d, Groupby)
mn = sim_brain_data.aggregate(mask, "mean")
assert isinstance(mn, Brain_Data)
assert len(mn.shape()) == 1
def series_to_brain(data, mask_x):
'''Converts a pandas series of ROIs to a Brain_Data instance. Index must correspond to ROI index'''
if not isinstance(data, pd.Series):
raise ValueError('Data must be a pandas series')
if len(mask_x) != len(data):
raise ValueError(
'Data must have the same number of rows as mask has ROIs.')
return Brain_Data([mask_x[x] * data[x] for x in data.keys()]).sum()
def check_brain_data(data):
'''Check if data is a Brain_Data Instance.'''
from nltools.data import Brain_Data
if not isinstance(data, Brain_Data):
if isinstance(data, nib.Nifti1Image):
data = Brain_Data(data)
else:
raise ValueError("Make sure data is a Brain_Data instance.")
return data
def test_indexing(sim_brain_data):
index = [0, 3, 1]
assert len(sim_brain_data[index]) == len(index)
index = range(4)
assert len(sim_brain_data[index]) == len(index)
index = sim_brain_data.Y == 1
assert len(sim_brain_data[index.values.flatten()]) == index.values.sum()
assert len(sim_brain_data[index]) == index.values.sum()
assert len(sim_brain_data[:3]) == 3
d = sim_brain_data.to_nifti()
assert d.shape[0:3] == shape_3d
assert Brain_Data(d)
def simulate_data(n_observations, y, p, sigma, mask):
''' Simulate Brain Data
Args:
n_observations: (int) number of data points
y: (array) one dimensional array of signal
p: (float) probability of signal in voxels
sigma: (float) amount of gaussian noise to add
Returns:
data: (list) of Brain_Data objects
'''
dat = Brain_Data(mask).apply_mask(mask)
new_data = np.zeros((dat.shape()[0], n_observations))
for i in np.where(dat.data == 1)[0]:
if np.random.randint(0, high=10) < p:
new_data[i, :] = y
noise = np.random.randn(new_data.shape[0], n_observations) * sigma
dat.data = (new_data + noise).T
return dat
def get_trialtype_pain_regressors(self,nifti_data,onset_file):
print("importing nifti")
#import the nifti
if (os.path.isfile(nifti_data + "nltoolstandard.nii.gz")):
msmrl1 = Brain_Data(
nifti_data + "nltoolstandard.nii.gz")
else:
msmrl1 = Brain_Data(
nifti_data + ".nii.gz")
msmrl1.write(nifti_data + "nltoolstandard.nii.gz")
#preprocess the nifti?
print("importing onsets")
#import the onset
onsets = onsets_to_dm(
onset_file,
TR=2,
runLength=msmrl1.shape()[0]
)
#process the onset files
#
onsets.sampling_rate=2
onsets_convolved=onsets.convolve()
for c in onsets_convolved.columns:
if sum(onsets_convolved.ix[:, c]) <= 0:
print('deleting '+ str(c))
del onsets_convolved[c]
onsets_convolved['linearterm']=range(1,361)
onsets_convolved['quadraticterm']=[pow(x,2) for x in onsets_convolved['linearterm']]
onsets_convolved['cubicterm']=[pow(x,3) for x in onsets_convolved['linearterm']]
onsets_convolved['ones']=[1]*360
msmrl1.X=onsets_convolved
print("convolved onsets; regressing...")
#regress
regression=msmrl1.regress()
print("Regressing; calculating similarity...")
msm_predicted_pain = regression['beta'].similarity(self.stats['weight_map'], 'dot_product')
onset_colnames = onsets_convolved.columns.tolist()
msm_predicted_pain_dict={}
for i, b in enumerate(msm_predicted_pain):
msm_predicted_pain_dict[onset_colnames[i]] = b
return msm_predicted_pain_dict
def test_predict_multi():
# Simulate data 100 images worth
sim = Simulator()
sigma = 1
y = [0, 1]
n_reps = 50
output_dir = '.'
dat = sim.create_data(y, sigma, reps=n_reps, output_dir=output_dir)
y = pd.read_csv('y.csv', header=None, index_col=None)
dat = Brain_Data('data.nii.gz', Y=y)
# Predict within given ROIs
# Generate some "rois" (in reality non-contiguous, but also not overlapping)
roi_1 = dat[0].copy()
roi_1.data = np.zeros_like(roi_1.data, dtype=bool)
roi_2 = roi_1.copy()
roi_3 = roi_1.copy()
idx = np.random.choice(range(roi_1.shape()[-1]), size=9999, replace=False)
roi_1.data[idx[:3333]] = 1
roi_2.data[idx[3333:6666]] = 1
roi_3.data[idx[6666:]] = 1
rois = roi_1.append(roi_2).append(roi_3)
# Load in all 50 rois so we can "insert" signal into the first one
# rois = expand_mask(Brain_Data(os.path.join(get_resource_path(), 'k50.nii.gz')))
# roi = rois[0]
from sklearn.datasets import make_classification
X, Y = make_classification(n_samples=100, n_features=rois[0].data.sum(), n_informative=500, n_redundant=5, n_classes=2)
dat.data[:, rois[0].data.astype(bool)] = X
dat.Y = pd.Series(Y)
out = dat.predict_multi(algorithm='svm', cv_dict={'type': 'kfolds', 'n_folds': 3}, method='rois', n_jobs=-1, rois=rois[:3], kernel='linear')
assert len(out) == 3
assert np.sum([elem['weight_map'].data.shape for elem in out]) == rois.data.sum()
# Searchlight
roi_mask = rois[:2].sum()
out = dat.predict_multi(algorithm='svm', cv_dict={'type': 'kfolds', 'n_folds': 3}, method='searchlight', radius=4, verbose=50, n_jobs=-1, process_mask=roi_mask)
assert len(np.nonzero(out.data)[0]) == len(np.nonzero(roi_mask.data)[0])
def expand_mask(mask):
""" expand a mask with multiple integers into separate binary masks
Args:
mask: nibabel or Brain_Data instance
Returns:
out: Brain_Data instance of multiple binary masks
"""
from nltools.data import Brain_Data
if isinstance(mask,nib.Nifti1Image):
mask = Brain_Data(mask)
if not isinstance(mask,Brain_Data):
raise ValueError('Make sure mask is a nibabel or Brain_Data instance.')
mask.data = mask.data.astype(int)
tmp = []
for i in np.unique(mask.data):
tmp.append((mask.data==i)*1)
out = mask.empty()
out.data = np.array(tmp)
return out
def expand_mask(mask):
""" expand a mask with multiple integers into separate binary masks
Args:
mask: nibabel or Brain_Data instance
Returns:
out: Brain_Data instance of multiple binary masks
"""
from nltools.data import Brain_Data
if isinstance(mask, nib.Nifti1Image):
mask = Brain_Data(mask)
if not isinstance(mask, Brain_Data):
raise ValueError('Make sure mask is a nibabel or Brain_Data instance.')
mask.data = mask.data.astype(int)
tmp = []
for i in np.unique(mask.data):
tmp.append((mask.data == i) * 1)
out = mask.empty()
out.data = np.array(tmp)
return out
def get_wager_nps_map(self, nps_map_filepath=None, data_mask=None):
using_custom_filepath = False
if nps_map_filepath is not None:
using_custom_filepath = True
else:
nps_map_filepath = self.nps_map_filepath
if (os.path.isfile(nps_map_filepath)):
nps = Brain_Data(nps_map_filepath, mask=data_mask)
self.decoder = nps
if using_custom_filepath:
self.decoder_origin = nps_map_filepath
else:
self.decoder_origin = 'nps'
else:
raise Exception("error; cannot find NPS map" + nps_map_filepath)
def test_data(tmpdir):
sim = Simulator()
r = 10
sigma = 1
y = [0, 1]
n_reps = 3
output_dir = str(tmpdir)
sim.create_data(y, sigma, reps=n_reps, output_dir=output_dir)
shape_3d = (91, 109, 91)
shape_2d = (6, 238955)
y=pd.read_csv(os.path.join(str(tmpdir.join('y.csv'))),header=None,index_col=None).T
flist = glob.glob(str(tmpdir.join('centered*.nii.gz')))
dat = Brain_Data(data=flist,Y=y)
# Test shape
assert dat.shape() == shape_2d
# Test Mean
assert dat.mean().shape()[0] == shape_2d[1]
# Test Std
assert dat.std().shape()[0] == shape_2d[1]
# Test to_nifti
d = dat.to_nifti
assert d().shape[0:3] == shape_3d
# # Test T-test
out = dat.ttest()
assert out['t'].shape()[0]==shape_2d[1]
# Test Regress
dat.X = pd.DataFrame({'Intercept':np.ones(len(dat.Y)),'X1':np.array(dat.Y).flatten()},index=None)
out = dat.regress()
assert out['beta'].shape() == (2,shape_2d[1])
# Test indexing
assert out['t'][1].shape()[0] == shape_2d[1]
# Test threshold
i=1
tt = threshold(out['t'][i], out['p'][i], threshold_dict={'fdr':.05})
assert tt.shape()[0] == shape_2d[1]
def test_extract_roi(sim_brain_data):
mask = create_sphere([12, 10, -8], radius=10)
assert len(sim_brain_data.extract_roi(mask, metric="mean")) == shape_2d[0]
assert len(sim_brain_data.extract_roi(mask,
metric="median")) == shape_2d[0]
n_components = 2
assert sim_brain_data.extract_roi(
mask, metric="pca",
n_components=n_components).shape == (n_components, shape_2d[0])
with pytest.raises(NotImplementedError):
sim_brain_data.extract_roi(mask, metric="p")
assert isinstance(sim_brain_data[0].extract_roi(mask, metric="mean"),
(float, np.floating))
assert isinstance(sim_brain_data[0].extract_roi(mask, metric="median"),
(float, np.floating))
with pytest.raises(ValueError):
sim_brain_data[0].extract_roi(mask, metric="pca")
with pytest.raises(NotImplementedError):
sim_brain_data[0].extract_roi(mask, metric="p")
s1 = create_sphere([15, 10, -8], radius=10)
s2 = create_sphere([-15, 10, -8], radius=10)
s3 = create_sphere([0, -15, -8], radius=10)
masks = Brain_Data([s1, s2, s3])
mask = roi_to_brain([1, 2, 3], masks)
assert len(sim_brain_data[0].extract_roi(mask,
metric="mean")) == len(masks)
assert len(sim_brain_data[0].extract_roi(mask,
metric="median")) == len(masks)
assert sim_brain_data.extract_roi(mask, metric="mean").shape == (
len(masks),
shape_2d[0],
)
assert sim_brain_data.extract_roi(mask, metric="median").shape == (
len(masks),
shape_2d[0],
)
assert len(
sim_brain_data.extract_roi(mask,
metric="pca",
n_components=n_components)) == len(masks)
def fetch_emotion_ratings(data_dir=None, resume=True, verbose=1):
"""Download and loads emotion rating dataset from neurovault
Args:
data_dir: (string, optional). Path of the data directory. Used to force data storage in a specified location. Default: None
Returns:
out: (Brain_Data) Brain_Data object with downloaded data. X=metadata
"""
collection = 1964
dataset_name = "chang2015_emotion_ratings"
data_dir = _get_dataset_dir(dataset_name,
data_dir=data_dir,
verbose=verbose)
metadata, files = download_collection(collection=collection,
data_dir=data_dir,
resume=resume,
verbose=verbose)
return Brain_Data(data=files, X=metadata)
def multi_threshold(t_map, p_map, thresh):
""" Threshold test image by multiple p-value from p image
Args:
stat: Brain_Data instance of arbitrary statistic metric
(e.g., beta, t, etc)
p: Brain_data instance of p-values
threshold: list of p-values to threshold stat image
Returns:
out: Thresholded Brain_Data instance
"""
from nltools.data import Brain_Data
if not isinstance(t_map, Brain_Data):
raise ValueError('Make sure stat is a Brain_Data instance')
if not isinstance(p_map, Brain_Data):
raise ValueError('Make sure p is a Brain_Data instance')
if not isinstance(thresh, list):
raise ValueError('Make sure thresh is a list of p-values')
affine = t_map.to_nifti().get_affine()
pos_out = np.zeros(t_map.to_nifti().shape)
neg_out = deepcopy(pos_out)
for thr in thresh:
t = threshold(t_map, p_map, thr=thr)
t_pos = deepcopy(t)
t_pos.data = np.zeros(len(t_pos.data))
t_neg = deepcopy(t_pos)
t_pos.data[t.data > 0] = 1
t_neg.data[t.data < 0] = 1
pos_out = pos_out + t_pos.to_nifti().get_data()
neg_out = neg_out + t_neg.to_nifti().get_data()
pos_out = pos_out + neg_out * -1
return Brain_Data(nib.Nifti1Image(pos_out, affine))
def train_model(image_list, algorithm, cross_validation, output_dir,
file_path_key='resampled_file', mask=None):
"""
:param image_list: A list of dictionaries of the form
{
'collection_id': '504',
'filename': 'Pain_Subject_1_Low.nii.gz',
'target': '1',
'resampled_file': 'path/to/the/resampled/file.nii.gz',
'original_file': 'path/to/the/original/file.nii.gz'
}
"""
tic = time.time() # Start Timer
try:
holdout = [int(item['subject_id']) for item in image_list]
except KeyError:
holdout = None
if cross_validation:
if holdout:
cross_validation['subject_id'] = holdout
elif cross_validation['type'] == 'loso':
raise ValueError(
"subject_id is required for a LOSO cross validation.")
extra = {}
if algorithm in ('svr', 'svm'):
extra = {'kernel': 'linear'}
categorical_mapping = None
if algorithm in CLASSIFICATION_ALGORITHMS:
classes = {item['target'] for item in image_list}
assert len(classes) == 2, ('More than two classes. '
'Classification requires binary data.')
categorical_mapping = {cls: index for index, cls in enumerate(classes)}
for image in image_list:
image['target'] = categorical_mapping[image['target']]
Y = pd.DataFrame([float(item['target']) for item in image_list])
file_path_list = [item[file_path_key] for item in image_list]
dat = Brain_Data(data=file_path_list, Y=Y)
if mask:
log.info('Applying a mask')
nifti_mask = nb.load(mask)
dat = dat.apply_mask(nifti_mask)
output = dat.predict(algorithm=algorithm,
cv_dict=cross_validation,
plot=True,
**extra)
weightmap_filename = '%s_weightmap.nii.gz' % algorithm
output['weight_map'].write(os.path.join(output_dir, weightmap_filename))
log.info("Elapsed: %.2f seconds", (time.time() - tic)) # Stop timer
result = {'weightmap': weightmap_filename,
'intercept': float(output['intercept']),
'scatterplot': '%s_scatterplot.png ' % algorithm,
'stats': {key: output[key].tolist()
for key in ('Y', 'yfit_xval', 'yfit_all')
if key in output},
'categorical_mapping': categorical_mapping,
'summary': get_summary(output)}
if 'roc' in output:
result['roc'] = output['roc']
return result
