def mstate_preprocess(confobj, eeg, eeg_info_study_obj):
"""
Preprocesses (normalization option, GFP peak extraction, average referencing) EEG to prepare data for microstate computation.
Parameters
----------
confobj : object of type MstConfiguration
Contains the following attributes: subtract_column_mean_at_start (bool), debug (bool), use_gfp_peaks (bool), force_avgref (bool), set_gfp_all_1 (bool), use_smoothing (bool), gfp_type_smoothing (string),
smoothing_window (int), use_fancy_peaks (bool), method_GFPpeak (string), original_nr_of_maps (int), seed_number (int), max_number_of_iterations (int), ERP (bool), correspondance_cutoff (double):
eeg : array
Shape ntf*nch, conatains the EEG data the microstate analysis is to be computed on.
eeg_info_study_obj : object of type EegInfo
Contains the following attributes: nch (number of channels), tf (number of time frames per epoch), sf (sampling frequency), chlist (channel list)
Returns
-------
eeg: array
Preprocessed EEG.
gfp_peak_indices: list
List of indices of the EEG that qualify as global field power peaks.
gfp_curve: 1D array
Global field power for each time frame.
"""
#################
# 2.)a Subtract mean across columns (the mean map) [must be done for each 2 sec segment] (not done by default)
#################
if confobj.subtract_column_mean_at_start:
if confobj.debug:
print 'column mean subtracted, does yield different results from published algorithm'
eeg=subtract_column_mean_epochwise(eeg, eeg_info_study_obj.TF)
#################
# 3.) COMPUTE GFP (ln1 or ln2)
#################
gfp_curve = compute_gfp(eeg, confobj.method_GFPpeak)
if confobj.debug:
print 'GFP Curve computed'
#################
#4.) Compute GFP Peaks (if the whole EEG is taken (use_gfp_peaks = False) it just returns the indices for the whole EEG
#################
gfp_peak_indices, gfp_curve = compute_gfp_peaks(gfp_curve, confobj.use_gfp_peaks, confobj.use_smoothing, confobj.gfp_type_smoothing, confobj.smoothing_window, confobj.use_fancy_peaks)
#################
# 2.) AVG REF
#################
if confobj.force_avgref:
if confobj.debug:
print 'Forced average reference'
eeg=TK_norm(eeg, gfp_peak_indices, eeg_info_study_obj.nch)
#################
#5.) Set all Maps to GFP=1 (sets the maps for each timeframe to gfp=1)
#################
if confobj.set_gfp_all_1:
print "GFP all maps set to 1"
eeg = set_gfp_all_1(eeg, gfp_curve)
return eeg, gfp_peak_indices, gfp_curve
def modmaps_preprocess(confobj, eeg, eeg_info_study_obj):
"""
Preprocesses (normalization option, GFP peak extraction, average referencing) EEG to prepare data for modelmap computation.
Parameters
----------
confobj : object of type MstConfiguration
Contains the following attributes: subtract_column_mean_at_start (bool), debug (bool), use_gfp_peaks (bool), force_avgref (bool), set_gfp_all_1 (bool), use_smoothing (bool), gfp_type_smoothing (string),
smoothing_window (int), use_fancy_peaks (bool), method_GFPpeak (string), original_nr_of_maps (int), seed_number (int), max_number_of_iterations (int), ERP (bool), correspondance_cutoff (double):
eeg : array
Shape ntf*nch, conatains the EEG data the microstate modelmap analysis is to be computed on.
eeg_info_study_obj : object of type EegInfo
Contains the following attributes: nch (number of channels), tf (number of time frames per epoch), sf (sampling frequency), chlist (channel list)
Returns
-------
eeg: array
Preprocessed EEG.
gfp_peak_indices: list
List of indices of the EEG that qualify as global field power peaks.
gfp_curve: 1D array
Global field power for each time frame.
"""
#################
# 2.) Subtract mean across columns (the mean map) [must be done for each 2 sec segment] (not done by default)
#################
if confobj.subtract_column_mean_at_start:
if confobj.debug:
print('column mean subtracted, does yield different results from published algorithm')
eeg=subtract_column_mean_epochwise(eeg, eeg_info_study_obj.TF)
#################
# 3.) COMPUTE GFP (ln1 or ln2)
#################
gfp_curve = compute_gfp(eeg, confobj.method_GFPpeak)
if confobj.debug:
print('GFP Curve computed')
#numpy.savetxt("gfp_curve_{0}" .format(eeg[0,0]), gfp_curve)
#################
#4.) Compute GFP Peaks (if the whole EEG is taken (use_gfp_peaks = False) it just returns the indices for the whole EEG
#################
gfp_peak_indices_pre, gfp_curve = compute_gfp_peaks(gfp_curve, confobj.use_gfp_peaks, confobj.use_smoothing, confobj.gfp_type_smoothing, confobj.smoothing_window, confobj.use_fancy_peaks)
#correction for artifical peaks at epoch transitions
#excludes all peaks that are divisible by 512: e.g. 0, 512, 1024 [beginning of an epoch if tf=512] or that are 1 smaller than a divisible: e.g. 511, 1023,... [ends of epochs if tf=511]
gfp_peak_indices = numpy.array([x for x in gfp_peak_indices_pre if not x % eeg_info_study_obj.tf == 0 and not (x+1) % eeg_info_study_obj.tf == 0])
#make sure that there are at least as many gfp peaks as confobj.original_nr_of_maps
if not len(gfp_peak_indices)>=confobj.original_nr_of_maps:
#add correct error type
raise IndexError("Number of GFP peaks need to be larger than number of maps extracted. Number of maps extracted is set to:" + confobj.original_nr_of_maps + ". Only " + len(gfp_peak_indices) + " GFP Peak Indices are available for computation.")
if confobj.debug:
print('GFP Peak Indices identified')
#numpy.savetxt("gfp_peak_indices_{0}" .format(eeg[0,0]), gfp_peak_indices)
#numpy.savetxt("gfp_peaks_{0}" .format(eeg[0,0]), eeg[gfp_peak_indices])
#################
# 5.) AVG REF (done by default)
#################
if confobj.force_avgref:
eeg=TK_norm(eeg, gfp_peak_indices, eeg_info_study_obj.nch)
if confobj.debug:
print('Forced average reference')
#numpy.savetxt("eeg_TKnorm_{0}" .format(eeg[0,0]), eeg)
#################
# 6.) Set all Maps to GFP=1 (sets the maps for each timeframe to gfp=1) (not done by default)
#################
if confobj.set_gfp_all_1:
print("GFP all maps set to 1")
eeg = set_gfp_all_1(eeg, gfp_curve)
return eeg, gfp_peak_indices, gfp_curve
#################
# 6.) Compute vectornorm for all maps -- is done for meanmods as well ---> Test!
#################
normalize_maps(eeg, 'vector_norm_1')
return eeg, gfp_peak_indices, gfp_curve
def modmaps_preprocess(confobj, eeg, eeg_info_study_obj):
"""
Preprocesses (normalization option, GFP peak extraction, average referencing) EEG to prepare data for modelmap computation.
Parameters
----------
confobj : object of type MstConfiguration
Contains the following attributes: subtract_column_mean_at_start (bool), debug (bool), use_gfp_peaks (bool), force_avgref (bool), set_gfp_all_1 (bool), use_smoothing (bool), gfp_type_smoothing (string),
smoothing_window (int), use_fancy_peaks (bool), method_GFPpeak (string), original_nr_of_maps (int), seed_number (int), max_number_of_iterations (int), ERP (bool), correspondance_cutoff (double):
eeg : array
Shape ntf*nch, conatains the EEG data the microstate modelmap analysis is to be computed on.
eeg_info_study_obj : object of type EegInfo
Contains the following attributes: nch (number of channels), tf (number of time frames per epoch), sf (sampling frequency), chlist (channel list)
Returns
-------
eeg: array
Preprocessed EEG.
gfp_peak_indices: list
List of indices of the EEG that qualify as global field power peaks.
gfp_curve: 1D array
Global field power for each time frame.
"""
#################
# 2.) Subtract mean across columns (the mean map) [must be done for each 2 sec segment] (not done by default)
#################
if confobj.subtract_column_mean_at_start:
if confobj.debug:
print(
'column mean subtracted, does yield different results from published algorithm'
)
eeg = subtract_column_mean_epochwise(eeg, eeg_info_study_obj.TF)
#################
# 3.) COMPUTE GFP (ln1 or ln2)
#################
gfp_curve = compute_gfp(eeg, confobj.method_GFPpeak)
if confobj.debug:
print('GFP Curve computed')
#numpy.savetxt("gfp_curve_{0}" .format(eeg[0,0]), gfp_curve)
#################
#4.) Compute GFP Peaks (if the whole EEG is taken (use_gfp_peaks = False) it just returns the indices for the whole EEG
#################
gfp_peak_indices_pre, gfp_curve = compute_gfp_peaks(
gfp_curve, confobj.use_gfp_peaks, confobj.use_smoothing,
confobj.gfp_type_smoothing, confobj.smoothing_window,
confobj.use_fancy_peaks)
#correction for artifical peaks at epoch transitions
#excludes all peaks that are divisible by 512: e.g. 0, 512, 1024 [beginning of an epoch if tf=512] or that are 1 smaller than a divisible: e.g. 511, 1023,... [ends of epochs if tf=511]
gfp_peak_indices = numpy.array([
x for x in gfp_peak_indices_pre if not x %
eeg_info_study_obj.tf == 0 and not (x + 1) % eeg_info_study_obj.tf == 0
])
#make sure that there are at least as many gfp peaks as confobj.original_nr_of_maps
if not len(gfp_peak_indices) >= confobj.original_nr_of_maps:
#add correct error type
raise IndexError(
"Number of GFP peaks need to be larger than number of maps extracted. Number of maps extracted is set to:"
+ confobj.original_nr_of_maps + ". Only " + len(gfp_peak_indices) +
" GFP Peak Indices are available for computation.")
if confobj.debug:
print('GFP Peak Indices identified')
#numpy.savetxt("gfp_peak_indices_{0}" .format(eeg[0,0]), gfp_peak_indices)
#numpy.savetxt("gfp_peaks_{0}" .format(eeg[0,0]), eeg[gfp_peak_indices])
#################
# 5.) AVG REF (done by default)
#################
if confobj.force_avgref:
eeg = TK_norm(eeg, gfp_peak_indices, eeg_info_study_obj.nch)
if confobj.debug:
print('Forced average reference')
#numpy.savetxt("eeg_TKnorm_{0}" .format(eeg[0,0]), eeg)
#################
# 6.) Set all Maps to GFP=1 (sets the maps for each timeframe to gfp=1) (not done by default)
#################
if confobj.set_gfp_all_1:
print("GFP all maps set to 1")
eeg = set_gfp_all_1(eeg, gfp_curve)
return eeg, gfp_peak_indices, gfp_curve
#################
# 6.) Compute vectornorm for all maps -- is done for meanmods as well ---> Test!
#################
normalize_maps(eeg, 'vector_norm_1')
return eeg, gfp_peak_indices, gfp_curve
