def preprocessNormalizeDownsample(self,
df,
df_interictal,
fs,
fsds,
montage="bipolar",
prewhiten=True):
#% Preprocessing
data, data_ref, _, data_filt, channels = echobase.preprocess(
df, fs, fsds, montage=montage, prewhiten=prewhiten)
dataII, _, _, dataII_filt, channelsII = echobase.preprocess(
df_interictal, fs, fsds, montage=montage, prewhiten=prewhiten)
###DID NOT CHECK IF channels and channelsII are equal. echobase.preprocess may got rid of channels in one but not the other if there are any Nans.
print("\n\n\nNormalizing data")
#normalize
dataII_scaler = echomodel.scaleData(dataII_filt, dataII_filt)
data_scaler = echomodel.scaleData(data_filt, dataII_filt)
#get annotated segments of preprocessed data
#alter annotations to include time_step seconds beforehand
#downsample
print("\n\n\nDownsampling data")
dataII_scalerDS = self.downsample(dataII_scaler, fs, fsds)
data_scalerDS = self.downsample(data_scaler, fs, fsds)
return dataII_scaler, data_scaler, dataII_scalerDS, data_scalerDS, channels
patientsWithseizures = DataJson.get_patientsWithSeizuresAndInterictal()
#%% generate example EEG
i = 42
sub = np.array(patientsWithseizures["subject"])[i]
idKey = np.array(patientsWithseizures["idKey"])[i]
AssociatedInterictal = np.array(patientsWithseizures["AssociatedInterictal"])[i]
df, fs, ictalStartIndex, ictalStopIndex = DataJson.get_precitalIctalPostictal(sub, "Ictal", idKey, username, password, BIDS = BIDS, dataset= dataset, session = session, secondsBefore = secondsBeforeSpread, secondsAfter = secondsAfterSpread)
df_interictal, _ = DataJson.get_iEEGData(sub, "Interictal", AssociatedInterictal, username, password, BIDS = BIDS, dataset= dataset, session = session, startKey = "Start")
print("\n\n\nPreprocessing\n\n\n")
dataII_scaler, data_scaler, dataII_scalerDS, data_scalerDS, channels = DataJson.preprocessNormalizeDownsample(df, df_interictal, fs, fsds, montage = montage, prewhiten = prewhiten)
data, data_ref, _, _, channels = echobase.preprocess(df, fs, fsds, montage = montage, prewhiten = prewhiten)
data_plotFilt = echobase.elliptic_bandFilter(data_ref, int(fs))[0]
nsamp, nchan = df.shape
#c = [62,61,60,59,58,53,52,51,43,42,36,32,31,27,21,17,7,6,5]#car
c = [47,45,64, 63,62,59,40,33,27,26, 18, 17,16,6,5] #bipolar
#c.reverse()
c_copy = copy.deepcopy(c)
fs_plot = copy.deepcopy(fs)
data_scalerDSPlot = data_scalerDS[:,c]
i = 86#RID0309: 33-43 ; RID0278:19; RID0596: 86
RID = np.array(patientsWithseizures["subject"])[i]
idKey = np.array(patientsWithseizures["idKey"])[i]
AssociatedInterictal = np.array(patientsWithseizures["AssociatedInterictal"])[i]
#%%
df, fs, ictalStartIndex, ictalStopIndex = DataJson.get_precitalIctalPostictal(RID, "Ictal", idKey, username, password,BIDS = BIDS, dataset= datasetiEEG, session = session, secondsBefore = secondsBefore, secondsAfter = secondsAfter)
df_interictal, _ = DataJson.get_iEEGData(RID, "Interictal", AssociatedInterictal, username, password, BIDS = BIDS, dataset= datasetiEEG, session = session, startKey = "Start")
print("preprocessing")
dataII_scaler, data_scaler, dataII_scalerDS, data_scalerDS, channels = DataJson.preprocessNormalizeDownsample(df, df_interictal, fs, fsds, montage = montage)
data, data_avgref, data_ar, data_filt, channels = echobase.preprocess(df, fs, fsds, montage = montage)
data_scalerDSDS = DataJson.downsample(data_scaler, fs, 16)
data_arDSDS = DataJson.downsample(data_ar, fs, 16)
nsamp, nchan = data_filt.shape
# %% Plot eeg Optional
#DataJson.plot_eeg(data_scalerDSDS, 16, markers = [secondsBefore*16], dpi = 25, aspect=aspect*2, height=nchan/aspect/2)
#for plotting eeg of 595
#data_DSDS = DataJson.downsample(data_filt, fs, 16)
#DataJson.plot_eeg(data_DSDS, 16, markers = [secondsBefore*16], dpi = 25, aspect=aspect*2, height=nchan/aspect/2)
#%%Meaure seizure spread
#%%Get data
patientsWithseizures = DataJson.get_patientsWithSeizuresAndInterictal()
i = 0
RID = np.array(patientsWithseizures["subject"])[i]
idKey = np.array(patientsWithseizures["idKey"])[i]
AssociatedInterictal = np.array(patientsWithseizures["AssociatedInterictal"])[i]
df, fs, ictalStartIndex, ictalStopIndex = DataJson.get_precitalIctalPostictal(RID, "Ictal", idKey, username, password, fpath = fpath_EEG, secondsBefore = secondsBeforeSpread, secondsAfter = secondsAfterSpread)
df_interictal, _ = DataJson.get_iEEGData(RID, "Interictal", AssociatedInterictal, username, password, fpath = fpath_EEG, startKey = "Start")
print("preprocessing")
dataII_scaler, data_scaler, dataII_scalerDS, data_scalerDS = DataJson.preprocessNormalizeDownsample(df, df_interictal, fs, fsds)
data, data_avgref, data_ar, data_filt = echobase.preprocess(df, fs, fsds)
data_scalerDSDS = DataJson.downsample(data_scaler, fs, 16)
data_arDSDS = DataJson.downsample(data_ar, fs, 16)
nsamp, nchan = df.shape
DataJson.plot_eeg(data_scalerDSDS, 16, markers = [secondsBeforeSpread*16], nchan = nchan, dpi = 25, aspect=aspect*2, height=nchan/aspect/2)
#%%window to fit model
data_scalerDS_X = echomodel.overlapping_windows(data_scalerDS, time_step, skip)
windows, _, _ = data_scalerDS_X.shape
probWN = np.zeros(shape = (windows, nchan))
probCNN = np.zeros(shape = (windows, nchan))
probLSTM = np.zeros(shape = (windows, nchan))
def get_FunctionalConnectivity(self,
sub,
idKey,
username,
password,
BIDS,
dataset,
session,
functionalConnectivityPath=None,
secondsBefore=30,
secondsAfter=30,
startKey="EEC",
fsds=256,
montage="bipolar",
FCtype="pearson"):
#checking if file is saved. If true, then just pull from that file
if not functionalConnectivityPath == None:
FCname = utils.baseSplitext(
self.getEEGfileName(sub, "Ictal", idKey, BIDS, dataset,
session, secondsBefore, secondsAfter))[:-4]
fname = join(
functionalConnectivityPath, FCname +
f"functionalConnectivity_{FCtype}_{montage}_interictalPreictalIctalPostictal.pickle"
)
if utils.checkIfFileExists(fname, printBOOL=False):
FC = utils.open_pickle(fname)
return FC
else:
# get data
AssociatedInterictalidKey = self.get_associatedInterictal(
sub, idKey)
seizure, fs, ictalStartIndex, ictalStopIndex = self.get_precitalIctalPostictal(
sub,
"Ictal",
idKey,
username,
password,
BIDS,
dataset,
session,
secondsBefore=180,
secondsAfter=180,
load=True)
interictal, fs = self.get_iEEGData(sub,
"Interictal",
AssociatedInterictalidKey,
username,
password,
BIDS,
dataset,
session,
startKey="Start",
load=True)
print("\nPreprocessing data: Filtering and Downsampling")
###filtering and downsampling
if FCtype == "coherence":
#Get Not prewhitened data (coherence measurement should not be used with pre-whitened data)
_, _, _, seizureFilt, channels = echobase.preprocess(
seizure, fs, fs, montage=montage, prewhiten=False)
_, _, _, interictalFilt, _ = echobase.preprocess(
interictal, fs, fs, montage=montage, prewhiten=False)
else:
#Get prewhitened data
_, _, _, seizureFilt, channels = echobase.preprocess(
seizure, fs, fs, montage=montage, prewhiten=True)
_, _, _, interictalFilt, _ = echobase.preprocess(
interictal, fs, fs, montage=montage, prewhiten=True)
ictalStartIndexDS = int(ictalStartIndex * (fsds / fs))
ictalStopIndexDS = int(ictalStopIndex * (fsds / fs))
#downsample
seizureFiltDS = self.downsample(seizureFilt, fs, fsds)
interictalFiltDS = self.downsample(interictalFilt, fs, fsds)
data = [
interictalFiltDS, seizureFiltDS[:ictalStartIndexDS, :],
seizureFiltDS[ictalStartIndexDS:ictalStopIndexDS + 1, :],
seizureFiltDS[ictalStopIndexDS + 1:, :]
]
data_FC = []
if FCtype == "pearson":
for x in range(4):
print(f"\n\n {x}/3")
data_FC.append(
list(echobase.pearson_wrapper(data[x], fsds)
[0])) #ind at [0] because don't need the p-values
if FCtype == "crossCorrelation":
for x in range(4):
print(f"\n\n {x}/3")
data_FC.append(
list(echobase.crossCorrelation_wrapper(data[x], fsds)))
if FCtype == "coherence":
for x in range(4):
print(f"\n\n {x}/3")
data_FC.append(
list(echobase.coherence_wrapper(data[x], fsds)))
FC = [channels, data_FC]
utils.savePickle([channels, data_FC], fname)
return FC