def _load_test_data_from_file(self, patient, filename, preprocess=True):
"""
Loading single file test data
:param patient:
:param filename:
:return:
"""
assert ( filename.find('test'))
print "\nLoading test data for " + patient + filename
eeg_data_tmp = EEGData(filename)
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
eeg_data = eeg_data[0]
fs = eeg_data.sample_rate
# preprocessing
data = eeg_data.eeg_data
params = self.params
params['fs']=fs
### comment if no preprocessing
if preprocess==True:
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(data,params)
x = self.feature_extractor.extract(eeg_data)
self.features_test.append(np.hstack(x))
def load_train_data_from_file(patient_name, filename, params=None):
"""
Loading single file training data
filename: full path to .mat file
return: (EEG data Instance, y_seizure, y_early)
"""
#print "\nLoading train data for " + patient_name + filename
eeg_data_tmp = EEGData(filename)
# a list of Instance's
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
eeg_data = eeg_data[0]
# eeg_data is now an Instance
# determine labels based on filename
if filename.find('interictal') > -1:
y_seizure = 0
y_early = 0
elif eeg_data.latency < 15:
y_seizure = 1
y_early = 1
else:
y_seizure = 1
y_early = 0
fs = eeg_data.sample_rate
# preprocessing
data = eeg_data.eeg_data
params['fs'] = fs
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(
data, params)
return (eeg_data, y_seizure, y_early)
def _load_test_data_from_file(self, patient, filename, preprocess=True):
"""
Loading single file test data
:param patient:
:param filename:
:return:
"""
assert (filename.find('test'))
print("\nLoading test data for " + patient + filename)
eeg_data_tmp = EEGData(filename)
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
eeg_data = eeg_data[0]
fs = eeg_data.sample_rate
# preprocessing
data = eeg_data.eeg_data
params = self.params
params['fs'] = fs
### comment if no preprocessing
if preprocess == True:
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(
data, params)
x = self.feature_extractor.extract(eeg_data)
self.features_test.append(np.hstack(x))
def _load_training_data_from_file(self,
patient,
filename,
preprocess=True):
"""
Loading single file training data
:param patient:
:param filename:
:return:
"""
#print "\nLoading train data for " + patient + filename
eeg_data_tmp = EEGData(filename)
# a list of Instance's
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
# eeg_data is now an Instance
eeg_data = eeg_data[0]
if filename.find('interictal') > -1:
y_seizure = 0
y_early = 0
elif eeg_data.latency < 15:
y_seizure = 1
y_early = 1
else:
y_seizure = 1
y_early = 0
fs = eeg_data.sample_rate
# preprocessing
data = eeg_data.eeg_data
params = self.params
params['fs'] = fs
### comment if no preprocessing
if preprocess == True:
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(
data, params)
###
x = self.feature_extractor.extract(eeg_data)
self.features_train.append(np.hstack(x))
self.type_labels.append(y_seizure)
self.early_labels.append(y_early)
def load_test_data_from_file(patient_name, filename, params=None):
"""
Loading single file test data
:return: EEG data Instance (no labels returned)
"""
assert (filename.find('test'))
#print "\nLoading test data for " + patient_name + filename
eeg_data_tmp = EEGData(filename)
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
# an Instance
eeg_data = eeg_data[0]
fs = eeg_data.sample_rate
data = eeg_data.eeg_data
params['fs'] = fs
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(
data, params)
return eeg_data
def _load_training_data_from_file(self, patient, filename, preprocess=True):
"""
Loading single file training data
:param patient:
:param filename:
:return:
"""
#print "\nLoading train data for " + patient + filename
eeg_data_tmp = EEGData(filename)
# a list of Instance's
eeg_data = eeg_data_tmp.get_instances()
assert len(eeg_data) == 1
# eeg_data is now an Instance
eeg_data = eeg_data[0]
if filename.find('interictal') > -1:
y_seizure=0
y_early=0
elif eeg_data.latency < 15:
y_seizure=1
y_early=1
else:
y_seizure=1
y_early=0
fs = eeg_data.sample_rate
# preprocessing
data = eeg_data.eeg_data
params = self.params
params['fs']=fs
### comment if no preprocessing
if preprocess==True:
eeg_data.eeg_data = preprocessing.preprocess_multichannel_data(data,params)
###
x = self.feature_extractor.extract(eeg_data)
self.features_train.append(np.hstack(x))
self.type_labels.append(y_seizure)
self.early_labels.append(y_early)
def test_combination(self, fold=3, max_segments=-1):
"""
Test the predictors using features given by each feature_extractor
in feature_extractors on the data specified by the patient argument.
:param max_segments: maximum segments to load. -1 to use the number of
total segments available. Otherwise, all segments (ictal and interictal)
will be randomly subsampled without replacement.
return: an instance of FeaturesPredictsTable """
loader = SubjectEEGData(self._patient,
self._data_path,
use_cache=True,
max_train_segments=max_segments)
# a list of (Instance, y_seizure, y_early)'s
train_data = loader.get_train_data()
fs = train_data[0][0].sample_rate
# preprocessing.
#params = {'fs':fs,
# 'anti_alias_cutoff': 100.,
## 'anti_alias_width': 30.,
# 'anti_alias_attenuation' : 40,
# 'elec_noise_width' :3.,
# 'elec_noise_attenuation' : 60.0,
# 'elec_noise_cutoff' : [49.,51.]}
# list of preprocessed tuples
params = self.params
params['fs'] = fs
# for (x, y_seizure, y_early) in train_data:
# x.eeg_data = preprocessing.preprocess_multichannel_data(x.eeg_data, self.params)
#train_data2 = []
#for (x, y_seizure, y_early) in train_data:
# x.eeg_data = preprocessing.preprocess_multichannel_data(x.eeg_data, params)
# train_data2.append(((x, y_seizure, y_early)))
#train_data =train_data2
# pre-extract features
features = [
] # list of feature tuples. list length = len(self._feature_extractors)
Y_seizure = np.array(
[y_seizure for (x, y_seizure, y_early) in train_data])
Y_early = np.array([y_early for (x, y_seizure, y_early) in train_data])
skf_seizure = cross_validation.StratifiedKFold(Y_seizure, n_folds=fold)
skf_early = cross_validation.StratifiedKFold(Y_early, n_folds=fold)
for i, feature_extractor in enumerate(self._feature_extractors):
#print 'Extracting features with %s'%str(feature_extractor)
#Xlist = [feature_extractor.extract(x) for (x, y_seizure, y_early)
# in train_data]
Xlist = []
for (x, y_seizure, y_early) in train_data:
#print '---------'
#print x.eeg_data.shape
params['fs'] = x.sample_rate
x.eeg_data = preprocessing.preprocess_multichannel_data(
x.eeg_data, params)
feat = feature_extractor.extract(x)
#print x.eeg_data.shape
#print feat.shape
Xlist.append(feat)
#Tracer()()
# Xlist = list of ndarray's
#print len(Xlist), Xlist[0].shape,len(Xlist[0])
n = len(Xlist)
#d = len(Xlist[0])
d = Xlist[0].shape[0]
# make 2d numpy array
#print n,d
X = np.zeros((n, d))
#print X.shape
for i in range(len(Xlist)):
#print Xlist[i].shape, X[i, :].shape
X[i, :] = Xlist[i].T
# chunk data for cross validation
# construct a list of 2d numpy arrays to be fed to XValidation
# tr_I = train index, te_I = test index
X_seizure = []
y_seizure = []
#Tracer()()
for tr_I, te_I in skf_seizure:
X_seizure.append(X[tr_I, :])
y_seizure.append(Y_seizure[tr_I])
X_early = []
y_early = []
for tr_I, te_I in skf_early:
X_early.append(X[tr_I, :])
y_early.append(Y_early[tr_I])
features.append((X_seizure, y_seizure, X_early, y_early))
T = CVFeaturesPredictorsTester.test_all_combinations(
features, self._feature_extractors, self._predictors)
return T
def test_combination(self, fold=3, max_segments=-1):
"""
Test the predictors using features given by each feature_extractor
in feature_extractors on the data specified by the patient argument.
:param max_segments: maximum segments to load. -1 to use the number of
total segments available. Otherwise, all segments (ictal and interictal)
will be randomly subsampled without replacement.
return: an instance of FeaturesPredictsTable """
loader = SubjectEEGData(self._patient, self._data_path, use_cache=True,
max_train_segments=max_segments)
# a list of (Instance, y_seizure, y_early)'s
train_data = loader.get_train_data()
fs = train_data[0][0].sample_rate
# preprocessing.
#params = {'fs':fs,
# 'anti_alias_cutoff': 100.,
## 'anti_alias_width': 30.,
# 'anti_alias_attenuation' : 40,
# 'elec_noise_width' :3.,
# 'elec_noise_attenuation' : 60.0,
# 'elec_noise_cutoff' : [49.,51.]}
# list of preprocessed tuples
params = self.params
params['fs']=fs
# for (x, y_seizure, y_early) in train_data:
# x.eeg_data = preprocessing.preprocess_multichannel_data(x.eeg_data, self.params)
#train_data2 = []
#for (x, y_seizure, y_early) in train_data:
# x.eeg_data = preprocessing.preprocess_multichannel_data(x.eeg_data, params)
# train_data2.append(((x, y_seizure, y_early)))
#train_data =train_data2
# pre-extract features
features = [] # list of feature tuples. list length = len(self._feature_extractors)
Y_seizure = np.array([y_seizure for (x, y_seizure, y_early) in train_data])
Y_early = np.array([y_early for (x, y_seizure, y_early) in train_data])
skf_seizure = cross_validation.StratifiedKFold(Y_seizure, n_folds=fold)
skf_early = cross_validation.StratifiedKFold(Y_early, n_folds=fold)
for i, feature_extractor in enumerate(self._feature_extractors):
#print 'Extracting features with %s'%str(feature_extractor)
#Xlist = [feature_extractor.extract(x) for (x, y_seizure, y_early)
# in train_data]
Xlist = []
for (x, y_seizure, y_early) in train_data:
#print '---------'
#print x.eeg_data.shape
params['fs']=x.sample_rate
x.eeg_data = preprocessing.preprocess_multichannel_data(x.eeg_data, params)
feat =feature_extractor.extract(x)
#print x.eeg_data.shape
#print feat.shape
Xlist.append(feat)
#Tracer()()
# Xlist = list of ndarray's
#print len(Xlist), Xlist[0].shape,len(Xlist[0])
n = len(Xlist)
#d = len(Xlist[0])
d = Xlist[0].shape[0]
# make 2d numpy array
#print n,d
X = np.zeros((n, d))
#print X.shape
for i in xrange(len(Xlist)):
#print Xlist[i].shape, X[i, :].shape
X[i, :] = Xlist[i].T
# chunk data for cross validation
# construct a list of 2d numpy arrays to be fed to XValidation
# tr_I = train index, te_I = test index
X_seizure = []
y_seizure = []
#Tracer()()
for tr_I, te_I in skf_seizure:
X_seizure.append(X[tr_I, :])
y_seizure.append(Y_seizure[tr_I])
X_early = []
y_early = []
for tr_I, te_I in skf_early:
X_early.append(X[tr_I, :])
y_early.append(Y_early[tr_I])
features.append( (X_seizure, y_seizure, X_early, y_early) )
T = CVFeaturesPredictorsTester.test_all_combinations(features,
self._feature_extractors, self._predictors)
return T